Bioelectric Medicine

Established Two different sciences answer to the name bioelectric medicine, and merging them is the commonest error in coverage of either. Bioelectronic medicine asks whether stimulating a peripheral nerve can do the work of a drug. Its founding observation is the inflammatory reflex, characterised by Kevin Tracey and colleagues from the 1990s: electrical or pharmacological stimulation of the vagus nerve inhibits cytokine release. Its instruments are implanted pulse generators, its evidence standard is the randomised controlled trial, and its regulator is the FDA. Frontier Developmental bioelectricity asks something else entirely: whether the standing voltage across a tissue carries instructive information about what shape that tissue should be. That is Michael Levin's programme, and its flagship is a flatworm that regenerates with two heads, permanently, after a transient perturbation of gap-junctional communication and with no change to its genome.

Established They share the word and almost nothing else — different mechanisms, different timescales, different journals, different communities, different evidence standards. This brief keeps them apart until the closing sections, where the only interesting question is whether they meet at all.

Established What changed in 2025 and 2026 is that the first field acquired a regulatory category and, for the first time, a properly controlled pivotal trial: an implanted vagus-nerve device for rheumatoid arthritis, approved in the United States, with a sham-controlled primary endpoint published in Nature Medicine. Frontier In the same window the second field produced Anthrobots — motile multicellular constructs that self-assemble from adult human lung cells with no genetic editing — and still has no independent replication of its central claim. Speculative The exotic proposition at the end of this brief is that bioelectric state is a control layer sitting above the genome, storing a target morphology that cells collectively work toward. That claim is falsifiable, under-tested, and would be enormous if it held.

Established The inflammatory reflex is real, mechanistically specified, and the strongest foundation anything in this brief rests on. Tracey's group established that stimulating the vagus nerve inhibits release of tumour necrosis factor and other cytokines, and that the circuit runs in both directions — the nervous system senses inflammation and modulates it. The canonical statement is Tracey's 2009 review in Nature Reviews Immunology. The mapped components include a projection from the vagal dorsal motor nucleus to the coeliac–superior mesenteric ganglion and a population of choline acetyltransferase-expressing T cells that release acetylcholine into the spleen. Established It is a neuroimmune circuit with named cell types, and everything the clinical field promises is an attempt to cash it into therapy.

Established Vagus nerve stimulation for epilepsy is the mature, quantified therapy that proves the delivery route works. A meta-analysis of 74 studies covering 3,321 patients reports an average 51 per cent reduction in seizure frequency at one year, roughly half of patients reaching at least a 50 per cent reduction at final follow-up, and about 8 per cent achieving complete seizure resolution. Frontier Those figures reach this brief through secondary literature rather than the meta-analysis itself; the Cochrane review of VNS for focal seizures is the higher evidence tier and the citation of record. Established The regulatory history is dated and unambiguous: an FDA neurological devices panel cleared VNS for epilepsy in 1997; treatment-resistant major depressive disorder followed in 2005; the non-invasive gammaCore device was cleared for episodic cluster headache in April 2017 and acute migraine in January 2018, and for hemicrania continua and paroxysmal hemicrania in 2021; Vivistim paired VNS for upper-extremity motor deficit after chronic ischaemic stroke was cleared in 2021. Established The therapeutic dose is coarse: a LivaNova depression implant defaults to about 1.25 mA at 20 Hz with a 250 microsecond pulse width, 30 seconds on every five minutes — a 10 per cent duty cycle, delivered to the left vagus because the right carries the cardiac branch. Established It is worth holding that number in mind whenever the word electroceutical implies precision dosing.

Established The harms are quantified too, and they are the reason electrode longevity rather than mechanism is the engineering story. A 25-year single-centre retrospective of 247 patients gives an overall surgical complication rate of 8.6 per cent — infection 2.6 per cent, haematoma 1.9 per cent, vocal cord palsy 1.4 per cent. Established Stimulation side effects at one year run to hoarseness in about 28 per cent and throat or chin paraesthesia in about 12 per cent; by year three the commonest complaint is shortness of breath at roughly 3.2 per cent. Frontier Sleep apnoea has been reported at 28 per cent in one small adult epilepsy series — a signal from a small sample, not a rate. Established None of this is disqualifying. It is the ordinary cost profile of an implanted lead on a cranial nerve, and it sets the bar any new indication has to clear.

Established RESET-RA is the single most important result in this field and its controlled effect is eleven percentage points. The pivotal trial of the SetPoint neuroimmune modulation system in rheumatoid arthritis is published: Tesser and colleagues, Nature Medicine 32:369–378, double-blind, randomised and sham-controlled, 242 patients in the intention-to-treat population (122 active, 120 sham), all of whom had already had an inadequate response to or intolerance of biologic or targeted synthetic DMARDs. Established The dose is a one-minute pulse train once daily at 10 Hz, up to 2.5 mA with an active-group mean of 1.8 mA, delivered automatically at night from a rechargeable device implanted on the left cervical vagus after four weeks of weekly titration; sham was 0 mA. Established The primary endpoint, ACR20 at three months, was 35.2 per cent active against 24.2 per cent sham: a difference of 11.8 points, 95 per cent confidence interval 0.6 to 23.1, P = 0.0209. It was met, and it was met narrowly — the lower bound of the interval is six-tenths of a point. Established Secondaries at three months: EULAR good or moderate response 60.7 per cent against 41.7 per cent (P = 0.0048); DAS28-CRP low disease activity or remission 26.1 per cent against 15.4 per cent (P = 0.0154); DAS28-CRP minimal clinically important difference 45.1 per cent against 32.5 per cent, which at P = 0.0528 did not survive multiplicity adjustment. Established Safety was good: no deaths, no unanticipated adverse device effects, a mean implant duration above 700 days, related serious adverse events in 1.6 per cent of patients — four events, all perioperative, all resolved. Non-serious implantation-related events ran to 15.6 per cent. Six patients elected explantation before twelve months, three of them for perceived lack of benefit.

Established The comparator that makes those numbers legible comes from the field's own founding human paper. Koopman and colleagues' 2016 first-in-human study in PNAS opens by noting that symptomatic relief “can be achieved in up to 50 per cent of patients using biological agents.” Established That study reported that an implanted nerve stimulator inhibited peripheral blood TNF, IL-1beta and IL-6, that stimulation up to four times daily significantly inhibited TNF production for 84 days, and that arthritis severity improved on standardised composite scores — and its abstract states neither a participant number nor a control arm, because the design was open-label. Established Read together, the honest summary is this: a device produced a small, statistically real benefit in a population that had already exhausted the drugs, at a 1.6 per cent serious-event rate. That is a genuine achievement and it is not drug substitution.

Established Deep brain stimulation is the calibration point for how large a neuromodulation effect can get. More than 244,000 patients had been implanted worldwide through 2019. FDA approvals run essential and Parkinsonian tremor 1997, Parkinson's disease 2002, dystonia 2003 under a humanitarian device exemption, obsessive-compulsive disorder 2009 likewise, epilepsy 2018. Frontier The commonly cited effect sizes — roughly 50 per cent improvement in UPDRS III in the dopamine-unresponsive off state for subthalamic targets, a 69 per cent reduction in OFF time, about a 50 per cent reduction in levodopa dose — and the commonly cited harms — 5 per cent intracranial haemorrhage, 8 per cent infection, 10 per cent lead migration, bilateral subthalamic speech impairment up to 33 per cent — both reach this brief through a tertiary aggregation, and several of the harm figures are implausibly wide. Treat them as orientation, not as citable rates. Established Adaptive DBS — sense a biomarker, titrate the stimulus — is FDA approved, in 2025. The approval is PMA P960009, supplement S478, covering “Model B35200 and Model B35300 Percept neurostimulators with BrainSense technology and adaptive Therapy”; an FDA Summary of Safety and Effectiveness Data exists for it and is cited by name and URL in the reference list of a 2026 npj Parkinson's Disease paper, and two independent 2025–2026 reviews list adaptive DBS for Parkinson's disease as FDA approved. Established The exact approval date could not be verified from the FDA's own database, which was unreachable; this brief therefore prints the year and no day. Frontier The evidence behind the approval is thinner than the approval implies. The headline randomised, blinded evidence is Oehrn and colleagues' crossover pilot in Nature Medicine: four male participants, at one centre, described by its own authors as establishing feasibility. Real-world performance is more sober still — Busch and colleagues report an open-label series of 8 patients, of whom 7 completed adaptive programming and 6 remained on it at four months; overall well-being improved at group level (P = 0.007) while general movement did not reach significance (P = 0.058), and 3 of 8 patients showed individually significant improvement. Frontier A clinical review puts the adaptive-DBS benefit at a 16.3 per cent decrease in bothersome symptoms at Level 3 evidence, and roughly 98 per cent of ADAPT-PD participants opted to continue on adaptive rather than revert to conventional stimulation. Preference is a real signal and it is not efficacy.

Established The developmental side has an uncontroversial core that critics of the strong claim do not dispute. Levin, Thorlin, Robinson, Nogi and Mercola showed in Cell in 2002 that asymmetries in H+/K+-ATPase and cell membrane potential constitute a very early step in left–right patterning. Adams, Masi and Levin showed in Development in 2007 that proton-pump-dependent changes in membrane voltage are necessary and sufficient to induce Xenopus tail regeneration — a strong, testable formulation and the high-water mark of the mainstream-accepted part of the programme. Membrane potential is an established regulator of proliferation and differentiation. Frontier The quantitative flatworm result exists and is worth quoting instead of the reviews: Emmons-Bell and colleagues, in International Journal of Molecular Sciences 16:27865, treated more than 243 fragments of genetically wild-type Girardia dorotocephala with a gap-junction blocker and obtained regenerated heads matching the anatomy of three other planarian species, with brain length-to-width ratios separating by ANOVA at P < 4.9 × 10^–14 and significant differences in anterior neoblast density. Frontier The flagship permanent-double-head result — Durant and colleagues, Biophysical Journal 112:2231 — reports planaria that regenerate as double-headed through subsequent rounds of amputation with an unmodified genome. Frontier Levin's own summary of the ectopic-eye work is that tadpoles carrying engineered eyes on their tails “can see quite well, despite the fact that the eyes connect to the spinal cord, not the brain.” Frontier Depolarisation has been reported to induce a metastatic-melanoma-like phenotype and hyperpolarisation to prevent oncogene-driven transformation. Speculative Every one of these results is real and published. Almost every one of them comes from a single laboratory, and that — not the absence of effect sizes, which do exist in the primary papers — is the serious objection.

Established Two large clinical categories sit inside this subject and this brief carries no numbers for them. Spinal cord stimulation for chronic pain and sacral neuromodulation for incontinence are, by implant volume, the largest deployed neuromodulation therapies in the world. Frontier The research behind this page reached no primary source for either, so no effect size or implant count is printed. That is a stated gap, and the first thing a next pass should close.

Established One access platform in this subject reaches deep brain targets with no hardware left behind, and it is the only route here that has repeatedly run a credible sham. Magnetic-resonance-guided focused ultrasound converges several hundred beams through the intact skull onto a focus a few millimetres across; at high intensity the focus coagulates tissue, and at low intensity it perturbs excitability without heating it. Established The ablative form holds the regulatory record: unilateral thalamotomy of the ventral intermediate nucleus was approved in the United States for medication-refractory essential tremor in 2016 on a randomised sham-controlled pivotal trial, with tremor-dominant Parkinson’s disease following in 2018 and staged second-side treatment later. Frontier The sham arm is the part this brief cares about. A patient can be positioned in the scanner, fixed in the frame and held there for the same duration with the transducer inert — a far stronger blind than any implanted stimulator can offer, and the reason this platform has produced controlled evidence at a rate the rest of the field has not.

Established What the platform buys with the absence of an implant is the loss of the thing an implant is for. A thalamotomy is a lesion: no dose, no duty cycle, no reprogramming visit and no explant, and this brief’s stimulator-longevity constraint simply vanishes, because nothing has been left in the patient to fail. Established Deep brain stimulation at the same target is adjustable and reversible and carries an implanted pulse generator with a finite battery. Frontier The two are therefore not competing therapies so much as opposite settlements of a single trade, and the ablative settlement is made once and cannot be revisited. Established The adverse-event profile follows from that: gait disturbance and paraesthesia are the common findings after thalamotomy, most resolve, and a minority persist at a year with no adjustment available. Frontier A brief built around a duty cycle of thirty seconds on in every five minutes should say plainly that the ultrasound route has no duty cycle at all.

Established The regulatory first is real and it is a category, not a product. An implanted vagus-nerve device is now approved in the United States for an autoimmune disease. That converts “electroceuticals” from a slogan into a thing a regulator has a file on, with a published sham-controlled pivotal trial behind it. Frontier It also converts the field's central question from “does the inflammatory reflex translate?” to “how large is the translated effect, and against what?” — which is a much harder question and one the approval does not answer.

Frontier The most consequential open question in this brief is whether a fifty-dollar ear clip does what a surgically implanted device does. A randomised, double-blind, sham-controlled trial of auricular transcutaneous vagus nerve stimulation in active rheumatoid arthritis exists — by Baker and colleagues, with a commentary published in Arthritis & Rheumatology in January 2024 — and the pack behind this brief could not reach it, because the publisher was unreachable. Frontier A 2026 review in Frontiers in Immunology describes auricular taVNS as having significantly increased ACR20 responder rates and reduced DAS28-CRP against sham stimulation, and separately describes a four-week bilateral protocol lowering DAS28-CRP with a substantial fall in circulating IL-6 and in ultrasound Power Doppler synovitis score. That review gives no sample sizes, no p-values and no ACR20 percentages, does not name the trial's authors, and never mentions the implanted device. Speculative So the state of knowledge is exactly this: the sham-controlled non-invasive trial exists, one secondary source calls it positive on ACR20, and no number from it can responsibly be printed. Speculative If the ear clip lands in the same effect-size neighbourhood as the implant's 11.8 points, that is a problem for the implant, not a vindication of the ear clip — because the implant's entire value proposition is that surgery buys you something the surface device cannot deliver. This brief flags that as unresolved and material.

Frontier Closed-loop is the shared frontier move on both the DBS and VNS sides, and it carries the same evidentiary weakness in both. Sensing a biomarker and titrating stimulation to it is obviously the right architecture; the difficulty is that every deployed system so far has been approved on feasibility-scale randomised evidence and characterised in the real world by small open-label series. Speculative For inflammation the sensing problem is harder than for Parkinson's disease, because there is no implant-readable analogue of the stimulation-entrained gamma oscillation that adaptive DBS uses as its biomarker.

Frontier Anthrobots are the developmental side's most striking recent object. Gumuskaya and colleagues report in Advanced Science that single cells derived from adult human lung self-construct, after about two weeks in extracellular matrix and without any genetic editing or manual sculpting, into motile multicellular constructs 30 to 500 microns across, moving on cilia at 5 to 50 microns per second in patterns ranging from tight loops to straight lines, with morphological class correlating significantly with movement type. They are reported as capable of traversing and inducing rapid repair of scratches in cultured human neural cell sheets in vitro. Frontier Two things follow. Unlike xenobots these are human, adult and genetically unmodified, which removes the frog-embryo distancing that let xenobots be filed under curiosity. And the neural repair result is scratch closure in a dish, not nerve regeneration — a promising assay outcome and nothing more.

Speculative Bioelectric control of tumour phenotype is the highest-value under-tested claim in the field. If depolarisation state is causally upstream of transformation, and if it is reachable with ion-channel modulators that already exist as approved drugs, the clinical prize is very large. The replication status is the problem: essentially all of it comes from one laboratory, in amphibians. Speculative A related and unfunded programme is the one Levin's own group points at — algorithmic models linking molecular-genetic data to morphogenesis, supported in part by DARPA Lifelong Learning Machines funding for a multiplexed microfluidic Xenopus culture system. Speculative That is where the programme is most testable: a model that prospectively predicts a novel anatomy from a designed bioelectric perturbation would settle the argument in one experiment.

Frontier Fascicle-selective stimulation is the frontier that would make the word electroceutical honest. The vagus is roughly 80 per cent afferent and carries fibres to many organs; a cuff electrode recruits a heterogeneous bundle. Selective recruitment of a defined fascicle, with closed-loop readout of the biomarker being modulated, is an active research area and has not reached clinical use. Frontier VNS in long COVID and post-viral inflammation is a live but very small literature — reduced inflammatory markers and fatigue in long COVID cohorts, modulated cytokine production and gut permeability after burn injury — and the sources describing it call several of the supporting studies very small.

Frontier The application with the largest downstream consequence is not a therapy but a door. Low-intensity focused ultrasound delivered with intravenous microbubbles transiently opens the blood–brain barrier over a targeted volume, and the opening closes within about a day. Established The first human demonstration in Alzheimer’s disease was a five-patient safety study reporting reproducible opening and closure without serious adverse events. Frontier The result that made the approach consequential paired the opening with an anti-amyloid antibody: in a three-participant study, regions treated with ultrasound alongside antibody infusion showed greater reduction in amyloid signal than the untreated homologous regions in the same patients over six months. Speculative Three patients and a within-subject comparison of an imaging biomarker is a mechanism demonstration and not evidence of clinical benefit, and this brief reports it as one. Frontier If it holds, the platform’s real product is delivery — antibodies, vectors, enzymes — and the neuromodulation story becomes the smaller half of it.

Frontier Low-intensity ultrasonic neuromodulation is the least settled of the three modes and carries the cleanest cautionary result anywhere in this brief. Sub-thermal pulses have been reported to modulate human sensory cortex and to reach deep targets no transcranial magnetic coil can address, with the mechanism usually attributed to mechanosensitive channels and still disputed. Established The cautionary result is that pulsed ultrasound produces an audible artefact by bone conduction, and rodent work showed that widespread cortical activation attributed to direct neuromodulation was substantially an indirect auditory response; deafened animals and smoothed pulse envelopes changed the effect. Frontier That is a sham-effect finding in the exact sense this brief uses the term, and it arrived from inside the field rather than from its critics. Speculative No human neuromodulation result obtained for this brief is controlled against an acoustically matched sham with the rigour the rodent work implies is necessary, and that is the cheapest thing the subfield could fix.

Established The two halves of this subject have different binding constraints, and neither is the one the field talks about most.

Speculative For bioelectric pattern rewriting, the ordered chain runs: (1) independent, blinded replication of the permanent-double-head planarian result by a laboratory with no stake in the framework; (2) direct longitudinal measurement of the two attractors — a bistable spatial voltage pattern in the same animals that predicts regenerated anatomy before it forms; (3) identification of a decoder, some molecular mechanism that transduces a spatial membrane-voltage pattern into transcriptional or mechanical output; (4) any instructive bioelectric result in a mammal; (5) a druggable handle, showing the effect is reachable with clinically available ion-channel modulators at tolerable systemic exposure; (6) a regulatory endpoint where changed anatomy is measurable and clinically meaningful. Speculative The binding link is (3), and it is a scientific unknown with no engineering or funding shortcut past it. Nobody knows what reads the pattern. Step (4) sits close behind as a second, independent unknown. Frontier Step (1) is the interesting one, because planaria are cheap and the experiment is undergraduate-scale. Its absence after nearly a decade is a pure incentive failure, and it is the single highest-leverage thing an institute could actually pay for.

Frontier For electroceuticals with drug-like specificity the chain runs: (1) complete functional maps of human vagal fascicles, validated against end-organ; (2) cuff or intraneural interfaces with a measured chronic selectivity index; (3) a biomarker the same implant can read, correlated with a validated inflammatory marker; (4) sham-controlled, adequately powered efficacy in one inflammatory indication; (5) head-to-head non-inferiority against an anti-TNF agent on the same endpoint, with cost and adverse-event comparison; (6) five-year lead survival with maintained effect. Established The binding link is (4), and it is institutional and commercial rather than scientific. The blunt version of the science already works well enough to obtain approval. What is missing is an evidence base a sceptical payer would accept, and the only party with an incentive to generate it is the party selling the device. Frontier This is the clearest case in the whole cluster of a field whose binding constraint is trial economics.

Established Beneath both chains sits hardware. Lead fracture, lead migration, insulation failure and fibrotic encapsulation set the real service life of every implant in this brief, and a therapy whose controlled benefit is eleven points cannot absorb a high revision rate. Frontier No published series gives five-year lead survival for a cervical vagus implant used at this duty cycle, so the durability half of the value proposition is currently an assumption. Speculative Note what is not on either critical path: mechanism. The inflammatory reflex is well enough characterised, and the developmental phenomena are well enough established, that neither field is waiting on more mechanism at the level it already has.

Established The rung that gates everything above it on this platform is the skull. Bone attenuates, aberrates and heats; a treatment plan corrects for it from a computed-tomography estimate of skull density, and candidates whose skull transmits poorly cannot be brought to ablative temperature and are screened out before treatment. Frontier The result is a therapy carrying an anatomical eligibility criterion unrelated to the disease, which is an unusual thing to find in a modern indication and is rarely stated in coverage of it. Established Geometry compounds it: deep central targets sit where the beams converge best and cortical targets sit in the near field where the skull is, so the platform is strongest exactly where an electrode is hardest to place. Speculative This brief could obtain no figure for the proportion of candidates excluded on skull transmission, and that proportion decides whether the platform is a general route or a selected one.

Frontier This subject waits on results that other fields would have to produce, and most of them are measurement problems. Speculative Developmental bioelectricity needs in vivo membrane-voltage imaging at tissue scale with enough spatial and temporal resolution to record a pattern rather than sample a point — voltage-sensitive dyes and genetically encoded voltage indicators are improving, and the programme's central claim is unfalsifiable without them. Speculative It needs ion-channel pharmacology with tissue selectivity, because the systemic exposure required to hold a tissue at a chosen potential is exactly the exposure that stops hearts. Speculative It needs a mammalian regeneration model in which anatomy is scored quantitatively and blind, because amphibian results have not crossed and nobody has shown why not.

Frontier Bioelectronic medicine depends on peripheral-nerve anatomy at fascicle resolution in humans, which is an anatomical and imaging programme rather than a device programme. Frontier It depends on chronic implantable biosensing for an inflammatory analyte — there is no equivalent of a continuous glucose monitor for TNF or CRP, and without one, closed-loop immunomodulation is open-loop with extra marketing. Frontier It depends on health-economic modelling that payers accept, which now exists in at least one formal cost-effectiveness analysis of the rheumatoid-arthritis device that this brief has on record but has not read. Established And both halves depend on something more mundane: trial infrastructure willing to run sham surgery, which is expensive, ethically fraught, and almost never funded by anyone without a device to sell.

Frontier One dependency runs the other way and is worth naming. The clinical half now depends on the developmental half staying separate from it in the public record. Frontier An approved implant with a sham-controlled trial behind it and a contested claim about voltage-encoded anatomy are being described in the same press vocabulary, and the reputational coupling is asymmetric: the implant gains nothing from the association and loses a great deal if the flatworm result fails to replicate in public. Speculative That is a communications dependency rather than a scientific one, and it is the sort of dependency fields usually discover only after the damage.

Speculative Six experiments would move this subject more than any amount of further review writing.

Speculative One: the cheap one. Pre-registered, blinded replication of the permanent-double-head planarian phenotype in an independent laboratory, under the published protocol. Measurement: proportion of animals showing the altered phenotype across successive amputation rounds, with the analysis plan fixed in advance. Cost is trivial; this is an institutional failure, not a scientific one. Speculative Two: the attractor. Longitudinal membrane-voltage imaging of the same animals across amputation cycles in one-headed and double-headed lines, showing two stable spatial patterns. Measurement: prospective predictive accuracy of regenerated morphology from the pre-regeneration voltage map. Speculative Three: the decoder. Screen for a mutant in which the voltage pattern is intact but the anatomical response is abolished. Measurement: a dissociation between measured Vmem pattern and regenerated anatomy. That result would identify the reader and end the underdetermination objection at a stroke. Speculative Four: the mammal. Any designed voltage perturbation producing a designed anatomical change in a mouse. Measurement: blind quantitative morphometry against a pre-registered predicted anatomy.

Frontier Five: the head-to-head nobody will pay for. Implanted cervical VNS against auricular taVNS against sham, in one inflammatory indication, powered for non-inferiority, run by a party with no device to sell. Measurement: ACR20 at three months in all three arms. Frontier This is the experiment that would settle the open question in section 3, and its result is genuinely unpredictable. Frontier Six: the specificity test. Show that stimulating one identified vagal fascicle produces the anti-inflammatory effect while stimulating others does not. Measurement: a chronic in vivo selectivity index paired with a cytokine readout. Speculative A negative result there would support the minority view that VNS acts through diffuse neuromodulation and that its correct comparator is a drug, not a wire.

Frontier Two supporting measurements would make the rest interpretable. A durability study reporting five-year lead survival and maintained effect for cervical vagus implants, which nobody has published and which every economic argument silently assumes. Frontier And an implant-derived signal correlated against a validated inflammatory marker in the same patients over months — not a demonstration that a device can record something, but a demonstration that what it records tracks disease. Speculative Until that second measurement exists, every closed-loop claim in the inflammatory space is an architecture diagram rather than a therapy.

Established The engineering requirement that actually binds is electrode longevity, and the trial data say so directly. RESET-RA reports a mean implant duration above 700 days with six elective explantations before twelve months, three of them for perceived lack of benefit. Established The 25-year VNS series gives an 8.6 per cent overall surgical complication rate. Deep brain stimulation, the mature comparator, carries lead migration and lead fracture rates in the high single digits to low tens of per cent in tertiary aggregations. Frontier A therapy whose controlled benefit is 11.8 percentage points has no headroom for a revision surgery.

Frontier The specific hardware requirements follow from the workback chain. A chronic cuff or intraneural interface with a measured selectivity index — target-fibre recruitment over off-target, in vivo, sustained over years, not measured acutely in an anaesthetised animal. Frontier A sensing channel on the same lead, because closed-loop requires a readout and adding a second implanted device doubles the surgical risk for a therapy with a small effect. Established A rechargeable power path that tolerates a one-minute-per-day duty cycle, which is the one place this field has it easy: the RESET-RA dose is a single daily pulse train, so energy is not the constraint that it is in DBS. Frontier Encapsulation that survives a decade on a moving cranial nerve, which is the same unsolved materials problem that limits every chronic neural implant.

Speculative On the developmental side the engineering requirement is instrumentation rather than implants: spatially resolved, chronic, non-destructive voltage readout in living tissue, plus a way to write a specified voltage pattern with better spatial control than bath-applied channel blockers. Optogenetic ion-channel actuation is the obvious route and has not been driven to pattern-scale control in a regenerating animal. Speculative The write problem is the harder half. Every published perturbation in this literature is essentially global — a blocker in the water, an mRNA injection into an early blastomere — and a global perturbation cannot demonstrate that a specific pattern maps to a specific anatomy. Handwave A voltage projector for tissue, addressing thousands of independent points at cellular resolution over days, is the instrument the strong claim needs, and nobody is building one.

Established Deep brain stimulation is the adjacent technology that matters most, because it is the same engineering problem at a different target. More than 244,000 implants, five distinct FDA-approved indications, and now a closed-loop variant approved in 2025 — everything bioelectronic medicine wants to be, arrived at twenty years earlier. Its lessons transfer directly: approval precedes strong evidence, real-world performance underperforms trial performance, and lead hardware sets the ceiling.

Established Spinal cord stimulation and sacral neuromodulation are the volume business of this field and are largely absent from its rhetoric. Both are established, regulated, widely implanted therapies. This brief prints no figures for either because the research behind it reached no primary source; the omission is stated rather than papered over. Frontier Non-invasive transcutaneous auricular and cervical devices are the adjacent form factor with the largest addressable population by orders of magnitude, and the smallest evidence base per claim made for them.

Frontier Synthetic morphology is adjacent to the developmental half. Xenobots — sub-millimetre constructs of frog skin and cardiac muscle that walk, swim, push pellets, aggregate debris in swarms and heal lacerations — and Anthrobots are the same programme's applied edge: if you can specify a target morphology, you can build one. Speculative Ion-channel pharmacology in oncology is adjacent in the other direction, and it is the route by which a bioelectric result would become a drug rather than a device. Handwave One junction nobody works on: if tissue-level voltage patterns are readable and writable, and if cultured neural networks can be trained, then the natural interface to a wetware computer is bioelectric rather than synaptic. There is no paper joining those two literatures. It is proposed here as a workback target, not reported as a finding. Frontier Adjacent in a less comfortable direction: the wearable and wellness electronics industry, which shares this field's anatomical vocabulary, none of its evidence standards, and much of its consumer attention. Speculative Whether that adjacency is a distribution channel or a contamination risk is genuinely undecided, and the answer probably depends on the auricular trial that nobody in this brief has been able to read.

Established The evidence base in bioelectronic medicine is manufacturer-shaped, and everyone involved knows it. The pivotal rheumatoid-arthritis trial was run behind the company that makes the device; the founding human study was open-label; NICE's 2020 assessment of VNS for treatment-resistant depression found that evidence on its efficacy is limited in quality and recommended randomised controlled trials, fifteen years after the FDA approved it for that indication. Established Approval and evidence quality are different things, and the transatlantic divergence on the same device is the cleanest demonstration of that available anywhere in medicine.

Frontier Nobody funds the study the field most needs: a sham-controlled trial powered for a non-epilepsy inflammatory endpoint, run by a party with no device to sell. Sham control for an implanted device requires either sham surgery or an implanted-but-inactive arm, both expensive and both ethically contested. Frontier A formal cost-effectiveness analysis of the rheumatoid-arthritis device now exists in the health-economics literature, which is the beginning of a payer conversation rather than the end of one; Medicare and provincial coverage positions are not established by anything this brief can cite.

Frontier On the developmental side the institutional failure is narrower and more damning. The flagship result of a fifteen-year programme has not been independently replicated, and the replication would be cheap. No funder rewards it: confirming another laboratory's flatworm is not a grant, and refuting it is a career risk in a small field. Speculative An institute that wanted maximum epistemic return per dollar in this whole subject should buy that replication and pre-register it, whichever way it comes out. Speculative A second institutional gap: there is no target class, no biomarker and no trial design that an oncology review panel recognises for a bioelectric account of cancer, so the hypothesis with the largest potential prize is structurally unfundable rather than disproven.

Frontier A third requirement is a boundary somebody actually polices. Regulated implants and unregulated consumer stimulators currently make claims of the same grammatical shape about the same nerve, and no institution owns the distinction in public. Speculative The workable version is not a ban but a disclosure standard: any device invoking the inflammatory reflex states whether its evidence is sham-controlled, and at what effect size. Speculative That is cheap, it is enforceable through advertising rules rather than device law, and it is the only intervention in this brief that a body with no laboratory could implement on its own.

Frontier The platform’s institutional problem is that its most valuable use is neither a device nor a drug. An ablation is a device and an antibody is a drug; barrier opening that makes an approved drug work better in a targeted volume is a combination, and no source obtained for this brief describes a settled regulatory pathway for it. Established A therapy whose distribution is set by where the scanners already are will be reported as available long before it is reachable.

Established Sham surgery is the field's standing ethical problem and its standing methodological requirement at the same time. RESET-RA solved it by implanting every participant and setting the sham arm to 0 mA — which is defensible, and which also means 120 people underwent cervical surgery for no therapeutic intent. Established The trial's own harm record makes that concrete: one intraoperative pharyngeal perforation, one transient vocal-cord paresis with dysphagia, and vocal cord paresis in 4.5 per cent of implanted patients. Frontier Consent for a device with an 11.8-point advantage over sham has to say so in those words, and marketing that leads with 57.6 per cent does not.

Frontier Direct-to-consumer bioelectric devices are the reputational hazard. Non-invasive vagus stimulation is cheap, unregulated in most wellness channels, and makes claims that a cleared medical device could not. Speculative The field's own history is the warning: the mid-century bioelectric programme was annexed by frank pseudoscience, and the annexation is a large part of why the legitimate phenomena were dismissed along with the quackery for forty years.

Frontier Anthrobots raise a consent and moral-status question that xenobots did not. They are built from adult human somatic cells, without genetic modification, and they self-assemble into motile constructs with reproducible behavioural classes. Speculative No existing tissue-donation consent framework contemplates a donor's cells being reassembled into a moving multicellular entity with a behavioural repertoire, and no governance framework says who decides what may be built from them. Handwave The moral-status question sitting under that is currently unanswerable in any principled way, and the honest position is that the field has bought itself several years to think about it.

Frontier There is also an equity question the enthusiasm rarely reaches. The implanted form of this therapy costs a surgeon, an operating theatre and a device; the non-invasive form costs almost nothing. Speculative If the two turn out to be comparable, the ethical failure will have been an evidence failure — a decade of trial money spent on the expensive arm because that is where the reimbursement was. Speculative If they turn out not to be comparable, the same result licenses a global market in devices that do not work, sold on the reputation of one that does.

Speculative If the strong version of developmental bioelectricity is right, it adds a control layer above the genome, and that is a change in what biology is for. A genome would specify the parts; a bioelectric setpoint would specify the arrangement, and the arrangement would be editable without touching the parts. Speculative Regenerative medicine would stop being a stem-cell delivery problem and become a signalling problem — you would not build the organ, you would set the target and let the tissue build it. Handwave At the far end that is anatomy on demand, and it is worth saying plainly that no step of that chain has landed in a mammal.

Frontier The modest version is civilizationally significant too, and much more likely. A second therapeutic modality — electrical rather than pharmacological — for chronic inflammatory disease would matter most where drug supply chains are the constraint: a device implanted once, powered by induction, with no cold chain and no monthly biologic infusion, is a different health-system object from an anti-TNF agent. Frontier Against that: implants require surgeons, and surgical capacity is scarcer in most of the world than drugs are. Speculative The version that would actually scale is the non-invasive one, which is exactly the version whose evidence base is weakest and whose key trial this brief could not read.

Speculative The deepest implication is epistemic rather than technical: if pattern information can be stored in a physiological state rather than in a molecular sequence, then a century of assuming that biological information means sequence information has been reading one channel of a two-channel system. Handwave That would reorganise more than medicine. Heredity, development and evolution are all currently narrated in terms of what is written in sequence; a second, physiological store that persists across tissue turnover would require the narration to be rebuilt, and the rebuilt version would be strange. Frontier It is worth holding both possibilities at once. The claim is not absurd, the phenomena that motivate it are real and published, and the specific experiment that would settle it costs less than a week of a mid-sized clinical trial.

What these horizons track: the two fields resolve on different clocks, and the near-term entries are questions with existing experiments attached rather than forecasts.

• 10 yr: Frontier The auricular-versus-implanted question is settled one way or the other, because the trial already exists and only needs reading and replicating. Sham-controlled evidence appears in a second inflammatory indication or fails to. Adaptive DBS accumulates a real-world evidence base larger than eight patients, and the 16.3 per cent symptom figure either holds or shrinks. Payer decisions on the rheumatoid-arthritis implant arrive, and they will do more to set the field's trajectory than any laboratory result in the same window. The independent planarian replication either lands or its continued absence becomes the story about the field rather than a footnote to it. Frontier Expect focused-ultrasound barrier opening to be tested against a clinical rather than an imaging endpoint on this horizon, and expect the trial to be sized in tens rather than hundreds.
• 25 yr: Speculative Fascicle-selective closed-loop stimulation with an implant-readable inflammatory biomarker reaches clinical use, or the field concedes that VNS is diffuse neuromodulation delivered by wire and its comparator is a drug. Chronic lead survival, not mechanism, is what decides which. On the developmental side, a candidate molecular decoder for spatial voltage patterns is identified, or the informational interpretation is quietly retired while the phenomena — left-right patterning, tail regeneration, proliferation control — stay in the textbooks under other headings.
• 50 yr: Speculative Bioelectric setpoint manipulation appears as an adjunct in indications where changed anatomy is measurable and modest — scar-free wound closure, bone non-union — rather than in limb regeneration, which is not a viable first indication and never was. Electroceuticals with genuine drug-like specificity exist for two or three inflammatory conditions and are reimbursed, or the modality settles permanently into the niche it occupies now: an option for patients who have exhausted the drugs.
• 100 / 250+ yr: Handwave Target morphology becomes a writable specification: organs regenerated in situ by setting a bioelectric target rather than by transplanting cells, and constructed multicellular agents built from a patient's own somatic cells for repair tasks, descended from the Anthrobots of the 2020s. Every link in that chain is currently missing — no decoder, no mammalian result, no replication — and the brief says so cheerfully rather than hedging it. This is the horizon the Institute exists to build a workback plan to, and the plan begins with a flatworm experiment that costs almost nothing.

• Depends on Depends on measurement science more than on device engineering. The developmental half cannot progress without chronic, spatially resolved, non-destructive membrane-voltage imaging in living tissue, and without optogenetic or pharmacological actuation precise enough to write a specified pattern rather than shift a whole animal. The clinical half depends on human peripheral-nerve anatomy at fascicle resolution, on a chronic implantable biosensor for an inflammatory analyte, and on trial infrastructure willing to run sham-controlled surgical studies. None of these is a bioelectric problem; all of them are borrowed from imaging, materials science and clinical trial methodology.
• Requires (not on this map) Two constraints sit outside the science and neither is being addressed by the people who would benefit most. The first is industrial: chronic stimulating leads on cranial and peripheral nerves fail by fracture, migration, insulation breakdown and fibrotic encapsulation, and a therapy whose controlled benefit is 11.8 percentage points cannot absorb a revision surgery. Electrode longevity is not a research topic with a funder; it is a manufacturing and materials capability that has improved slowly and incrementally for thirty years, and it caps every clinical claim in this brief. The second is institutional: the field needs an enforced boundary between regulated implanted devices with sham-controlled evidence and the consumer bioelectric market, where the same anatomical vocabulary is used to sell ear clips on far weaker evidence. That boundary is not a matter of taste. The mid-century bioelectric programme was dismissed for forty years partly because its legitimate phenomena were indistinguishable, to a regulator or a physician, from the electromagnetic healing industry that grew up beside it. If a cleared rheumatoid-arthritis implant and an unregulated wellness gadget make claims of the same shape, the category collapses again, and the collapse takes the real results with it. A third constraint arrives with the focused-ultrasound platform and it is a supply constraint rather than a scientific one: an installed base of magnetic-resonance-guided transducer suites outside academic and quaternary centres. The equipment is a transducer helmet inside an MRI suite at capital cost in the millions, no research result reduces that cost, and a therapy distributed by where the scanners already sit is available in the literature long before it is reachable by a patient.
• Enables Enables, if the clinical half consolidates, a second therapeutic modality for chronic inflammatory disease that does not depend on a biologic supply chain, and a template for regulatory approval of devices whose mechanism is neuromodulatory rather than pharmacological. Enables, if the developmental half consolidates, a route to regenerative medicine that specifies outcomes rather than delivering cells, and a control interface for engineered multicellular constructs built from a patient's own tissue. The second of those is contingent on a scientific result nobody has produced; the first is contingent only on trials nobody wants to fund.
• Adjacent Adjacent to deep brain stimulation and the whole implanted-neuromodulation industry, which supplies the hardware, the surgical pathway and the cautionary evidence pattern. Adjacent to synthetic morphology — xenobots and Anthrobots — which is the same programme building rather than repairing. Adjacent to ion-channel pharmacology, which is how a bioelectric result would become a pill instead of a device. Adjacent, speculatively and with nobody working on the junction, to cultured-neuron computing, where a bioelectric rather than synaptic interface would be the natural control channel.

Established “The rheumatoid arthritis device gets half of patients into response.” No. The figures in circulation — 50 per cent at six months, 57.6 per cent ACR20 at twelve months, 77.3 per cent EULAR good or moderate response — are open-label all-completer numbers with no control arm behind them. Established The controlled result is ACR20 35.2 per cent against 24.2 per cent sham at three months: 11.8 points, with a confidence interval running from 0.6 to 23.1. The uncontrolled numbers are the ones that travel and they should not be quoted without the word open-label attached.

Established “Adaptive deep brain stimulation is not yet approved in the United States.” Stale and wrong. Medtronic's BrainSense adaptive DBS was approved by the FDA in 2025 under PMA P960009, supplement S478. Widely used reference pages still assert the opposite. Established This brief prints the year without a day, because the FDA's own database was unreachable and no published source reachable here gives a date. Frontier The related trap is the “40 per cent reduction in motor symptoms” figure attached to adaptive DBS in coverage: it traces to a four-person blinded pilot whose authors described it as a feasibility study, and a real-world series of eight patients found group-level movement improvement at P = 0.058 with three of eight individually improved.

Established “Bioelectricity is one field.” It is two. Compare the source sets: the developmental literature and the bioelectronic-medicine literature share no authors, no journals and no methods. Most popular coverage merges them, usually by putting Levin's flatworms in the same paragraph as a vagus implant.

Established “Levin showed that memories can be stored in body voltage.” The published claim is about anatomical pattern memory — the target shape a regenerating animal reconstructs — not declarative memory. The planarian memory-after-decapitation literature is separate and much more contested, and conflating the two does the stronger claim no favours.

Established “The two-headed planaria are genetically modified.” They are not, and that is the entire content of the claim. The perturbation is transient blockade of gap-junctional communication; the genome is unchanged; the phenotype persists through further rounds of amputation.

Established “Xenobots reproduce.” They perform kinematic self-replication: swarms push loose dissociated cells into piles that become new swarms. There is no genetic replication, no heredity and no autonomous reproduction. The word in the paper title is doing a great deal of work that the mechanism does not support.

Established “Vagus nerve stimulation is a targeted immune switch.” Clinical VNS delivers on the order of 1.25 mA at 20 Hz on a 10 per cent duty cycle to a mixed nerve that is roughly 80 per cent afferent. Fibre-selective stimulation is a research goal, not a product. Frontier A serious minority position holds that the therapeutic action may be entirely non-specific — chronic mild afferent stimulation raises noradrenaline from locus coeruleus, serotonin from dorsal raphe and dopamine in prefrontal cortex and nucleus accumbens — in which case bioelectronic medicine is a delivery route for diffuse neuromodulation and its correct comparator is a drug. That view is refutable by the fascicle-selectivity experiment in section 6, and the experiment has not been run.

Established “VNS is proven for depression.” It has been FDA-approved for treatment-resistant depression since 2005, and NICE concluded in 2020 that evidence on its efficacy is limited in quality and recommended randomised controlled trials. Both statements are true simultaneously; approval is not proof.

Frontier “The ear clip gives you what the implant gives you” — and its mirror image, “the ear clip is obviously worthless.” Neither can be asserted. A randomised, double-blind, sham-controlled trial of auricular VNS in active rheumatoid arthritis exists; the pack behind this brief could not open it because the publisher was unreachable; a 2026 review describes it as positive on ACR20 without giving a sample size, a p-value or a percentage. Speculative This brief flags it as the field's most consequential unresolved question, and notes the direction that matters: if the non-invasive form matches the implant's eleven-point effect, the implant has a value problem, not the ear clip.

Frontier Sceptic-side: “bioelectric developmental claims are fringe.” They are not. The 2002 Cell paper on H+/K+-ATPase asymmetry in left–right patterning and the 2007 Development paper on proton flux as necessary and sufficient for tail regeneration are mainstream, well-cited developmental biology. What is contested is the informational interpretation, not the phenomena.

Speculative And the strongest form of the sceptical case, stated properly: the effects are real and the interpretation is overreach. Voltage is coupled to nearly everything — pH, calcium, proliferation, gap-junction state, cytoskeletal tension, gene expression. Showing that perturbing voltage changes anatomy shows that voltage sits somewhere in a densely coupled causal web; it does not show that voltage carries the pattern information. Speculative An information-bearing code needs a reader, and no molecular mechanism has been identified that reads a spatial voltage pattern and outputs an anatomical target. Without one, “bioelectric code” describes a correlation. Speculative The vocabulary compounds it: goals, memory, Selves, cognitive light cone are doing argumentative work the data do not underwrite, and critics reasonably read that as naming the phenomenon after the conclusion. Frontier Developmental biology already has a working account of pattern formation — morphogen gradients, reaction-diffusion, gene regulatory networks — with molecular identities for every component, and Occam's razor is not a fallacy.

Speculative The steelman for Levin, which the Institute should hold with equal seriousness. Every one of those objections was made against the genetic code before the ribosome was found, and against the action potential before Hodgkin and Huxley. The programme has produced predictive, surprising, reversible interventions, which is the strongest evidence class developmental biology offers, and predictive success ahead of a known decoder is how biology usually proceeds. Frontier The existing morphogen account also does not explain scale invariance, regeneration to a target, or why perturbed embryos stop when the right shape is reached — and those are precisely the phenomena bioelectricity is invoked for. Speculative What would have to be true for the strong version: a molecular reader of spatial membrane-voltage patterns; prospective prediction of a novel anatomy from a designed voltage pattern; and independent replication of the planarian result. None of the three has landed. Speculative What would falsify it: a systematic failure of independent laboratories to reproduce the permanent-double-head phenotype under the published protocol. For the clinical half, the falsifier is a properly sham-controlled, adequately powered trial in an inflammatory indication showing no effect. Neither experiment has been run, and saying so is a stronger position than either advocacy or dismissal.

Frontier Enthusiast-side: “bioelectric medicine will let us regrow limbs.” There is no instructive bioelectric result in a mammal. The chain from flatworm to human runs through at least four unlanded links, and the absence of a mammalian result after two decades is itself the evidence.

Handwave The historical claim the Institute finds attractive, flagged as unsourced. H. S. Burr's L-fields in the 1930s and 1940s and Robert O. Becker's regeneration work from the 1960s made something close to the modern bioelectric claim decades early, were dismissed for irreproducible measurements and absent mechanism, and were rapidly annexed by electromagnetic healing quackery — while Becker's specific finding that injury currents are necessary for amphibian limb regeneration survived and is now mainstream. Handwave That reading is exactly the kind of case the Institute exists to examine, and the research behind this brief did not open a single Burr or Becker primary source. It is printed here as an argument to be checked, not as history established.