In 2012, at the University Clinical Center in Sarajevo, surgeons implanted a small electrical device onto the left cervical vagus nerve of a patient whose rheumatoid arthritis had stopped responding to every drug her rheumatologists had tried. The device looked almost identical to the vagus nerve stimulators that had been used for two decades to treat epilepsy, a silicone cuff wrapped around the nerve in the neck, wired to a pulse generator tucked under the skin of the chest. What was different was the intent. Nobody was trying to calm a seizure. The surgical team, working with researchers from the Academic Medical Center in Amsterdam and the Feinstein Institutes for Medical Research in New York, was testing a theory that the nervous system itself could act as a drug: that if you deliver electricity to the right nerve at the right dose, the body’s own circuitry for suppressing inflammation would switch on.
The theory had a name, the inflammatory reflex, and a primary author, Kevin Tracey, who had been chasing it since 1998, when his lab at the Feinstein Institutes noticed that cutting the vagus nerve in rats sent levels of the inflammatory molecule TNF-alpha soaring, while stimulating the intact nerve suppressed it. By 2016, that line of research had produced a peer-reviewed clinical result: a study published in the Proceedings of the National Academy of Sciences, led by rheumatologist Frieda Koopman and senior author Paul-Peter Tak of the Academic Medical Center, showed that stimulating the vagus nerve in rheumatoid arthritis patients lowered the production of inflammatory cytokines and improved their disease activity scores. It was the first time a nerve, rather than a drug, had been shown in a controlled clinical setting to turn down the immune response driving an autoimmune disease.
The Inflammatory Reflex Turns a Nerve Into a Drug Target
Rheumatoid arthritis is driven by cytokines, small signaling proteins that macrophages and other immune cells pump out to coordinate inflammation. Tumor necrosis factor, interleukin-6, and interleukin-1 beta are the main offenders, and the entire biologic drug industry, Humira, Enbrel, Actemra, exists to block them chemically. Tracey’s work proposed a different lever. His 2000 paper with Borovikova in Nature showed that electrically stimulating the vagus nerve in animals suppressed TNF production during endotoxin-induced shock, and that cutting the nerve removed the brake entirely.
Stimulate the vagus, dial down the inflammation. Cut the vagus, lose the brake.
The circuit turned out to be more specific than a simple nerve-to-organ wire. Work from Tracey’s group, including a 2011 Science paper by Mauricio Rosas-Ballina, mapped a pathway in which vagal efferent signals travel to the celiac ganglion and then along the splenic nerve, a noradrenergic pathway rather than the classic vagal neurotransmitter acetylcholine. Norepinephrine released in the spleen acts on a population of CD4-positive T cells that, unusually, synthesize and release acetylcholine themselves. That acetylcholine then binds the alpha-7 nicotinic receptor on splenic macrophages, shutting down the NF-kB signaling that drives transcription of TNF and other cytokines. The spleen, in this model, is not a passive filter of blood. It is a relay station in a neural circuit that the brain uses to police inflammation in real time.
That mechanistic specificity is what made rheumatoid arthritis, a disease defined by chronic overproduction of exactly these cytokines, an obvious test case. If the inflammatory reflex was real and not just a laboratory curiosity in rodents, stimulating the human vagus nerve should measurably suppress the same cytokines that biologic drugs are built to block, without a drug at all.
What the 2016 Trial Actually Measured
The Koopman study enrolled patients with active rheumatoid arthritis who had already failed standard treatment, implanted the Cyberonics-style vagus nerve stimulator used in the trial, and then did something that distinguished it from epilepsy or depression VNS protocols: it ran the device at very low intensity for only a minute or so at a time, once or several times daily, rather than delivering continuous stimulation. The dosing question, how little electricity is enough to trigger the reflex without causing side effects like bradycardia or voice changes from off-target nerve fibers, turned out to matter as much as whether stimulation worked at all.
It did work, at least in this small, open-label cohort. Blood drawn from stimulated patients and exposed to lipopolysaccharide in the lab, a standard way of provoking an immune response ex vivo, produced less TNF than blood drawn before treatment began. Clinical disease activity scores, the composite measures rheumatologists use to track joint swelling, tenderness, and inflammatory blood markers, improved in most of the patients who completed the study. Patients who had cycled through multiple failed biologics, people for whom Humira, Enbrel, or Actemra simply stopped working or never worked, saw their inflammatory markers move in the right direction from an implanted device rather than an injection.
A nerve, not a needle, moved the same markers biologic drugs are built to block.
The result mattered less for its size, this was a small pilot study, not a registration trial, and more for what it proved was possible: that a reflex arc discovered in rat sepsis models could be reproduced, measured, and clinically meaningful in human autoimmune disease. Co-authors on the paper, including SetPoint Medical scientists Michael Faltys and Yaakov Levine, had spent years building the implantable hardware specifically to make that translation work, a device small enough and precise enough to deliver the microampere-level doses the reflex apparently needs.
From Academic Pilot to Two Competing Industrial Bets
The Koopman trial did not stay an academic curiosity. SetPoint Medical, the company that supplied the implant and co-authored the paper, used it as the foundation for a larger, randomized, sham-controlled multicenter trial, the kind of study regulators actually require before a device can be marketed. That later trial tested the central claim under harder conditions: patients who did not know whether their implanted device was turned on, compared against patients who did know, to see whether the anti-inflammatory effect survived rigorous blinding.
SetPoint was not the only serious money chasing this idea. In 2016, GSK and Verily, Alphabet’s life sciences unit, formed a joint venture called Galvani Bioelectronics specifically to pursue “bioelectronic medicines,” implanted devices meant to treat chronic diseases including rheumatoid arthritis by modulating peripheral nerves rather than dosing the bloodstream with drugs. Paul-Peter Tak, the rheumatologist who co-led the original Amsterdam trial, later moved into GSK’s immunology leadership, a career move that traces almost exactly the industrial path this research took, from a university hospital in Amsterdam to a pharmaceutical company betting that nerves could replace some of its own injectable drug pipeline.
| Approach | Mechanism | Delivery | Systemic exposure |
|---|---|---|---|
| TNF inhibitors (Humira, Enbrel) | Antibody binds and neutralizes circulating TNF | Subcutaneous injection, weekly or biweekly | Whole body, continuous |
| IL-6 inhibitors (Actemra) | Antibody blocks IL-6 receptor signaling | Infusion or injection | Whole body, continuous |
| Vagus nerve stimulation (SetPoint) | Activates splenic cholinergic pathway, suppresses macrophage cytokine release | Implanted cuff electrode, minutes of stimulation per day | Localized nerve activation, systemic cytokine effect |
The Case for Caution Before Celebration
None of this makes vagus nerve stimulation a replacement for biologic drugs, at least not yet, and the obvious caveats deserve more weight than they usually get in coverage of bioelectronic medicine. Blinding a surgical implant trial is genuinely hard. Patients know they underwent surgery, physicians can often tell from heart rate changes whether a device is active, and sham-controlled device trials are structurally harder to run cleanly than a placebo pill trial. The original 2016 study was small and open-label for much of its course, which is exactly why the larger, blinded multicenter trial that followed matters more for regulatory purposes than the headline-grabbing pilot.
There is also the question of durability and scale. Biologic drugs now treat hundreds of thousands of rheumatoid arthritis patients with response rates and safety profiles built on over two decades of real-world data. An implanted nerve stimulator requires surgery, carries surgical risk, and has to prove not just that it works but that it works reliably enough, for long enough, to justify putting a device in someone’s neck instead of writing a prescription. Rheumatologists are conservative by training, for good reason. A mechanism discovered in rat endotoxin shock in 1998 has had to clear an unusually long chain of validation, from rodent sepsis to primate anatomy to a human pilot study to a blinded multicenter trial, before it could even be considered alongside a biologic.
What the research does establish, cleanly, is that the inflammatory reflex is not a laboratory artifact confined to anesthetized rats. It operates in conscious human patients with an actual autoimmune disease, it is measurable in their blood, and it tracks with how their joints feel. That is a narrower claim than “vagus nerve stimulation cures rheumatoid arthritis,” but it is a much stronger claim than most immune system theories ever manage to prove outside a mouse.
The open question is not whether a nerve can talk to the immune system. Koopman, Tracey, Tak, and the SetPoint and Galvani engineers who followed them have shown that it can. The open question is whether medicine is willing to treat a wire in the neck as a legitimate alternative to a syringe, and whether insurers, surgeons, and patients will make that same bet before the first wave of bioelectronic rheumatoid arthritis devices reaches a pharmacy counter that does not exist yet, because this one does not get filled, it gets implanted.
Credit: Piron Guillaume on Unsplash