In 2011, at the Academic Medical Center in Amsterdam, a rheumatologist named Frieda Koopman implanted a device into the neck of a patient who had run out of options. The patient had rheumatoid arthritis that had failed to respond to methotrexate and multiple biologic drugs, the kind of case rheumatologists describe, without much hope, as refractory. The device was not built for arthritis. It was a cervical vagus nerve stimulator, the same category of hardware used for decades to treat epilepsy, repurposed on the theory that the nerve it targets does more than regulate heart rate and digestion. It also talks to the immune system. Five years later, Koopman, working with Paul-Peter Tak and Kevin Tracey of the Feinstein Institutes, published results in the Proceedings of the National Academy of Sciences: in a small group of patients, stimulating the vagus nerve suppressed production of tumor necrosis factor, the same inflammatory molecule that biologic drugs like adalimumab are built to block, and disease activity scores improved enough that most of the cohort reached low disease activity or remission.
That 2016 paper has become one of the most cited pieces of evidence that bioelectronic medicine is not a metaphor. It is now getting a second act. Longer follow-up data out of the clinical program built around that original finding shows the effect holding up at twelve months, the longest duration reported so far for an implanted device used against an autoimmune disease. For a field that has spent two decades proving a mechanism exists, durability is the harder and more commercially relevant question. Mechanisms get you papers. Durability gets you a product.
The Nerve Was Never Just About Heart Rate
Tracey’s lab at the Feinstein Institutes spent the late 1990s and 2000s mapping what it now calls the inflammatory reflex: a wired circuit, not a chemical one, that the body uses to keep inflammation from running away from itself. Cut the vagus nerve in an animal model of sepsis and inflammatory cytokines spike. Stimulate it and they drop. The pathway runs through the spleen, where acetylcholine released by a population of T cells, described by Tracey’s group and collaborator Mauricio Rosas-Ballina in a 2011 Science paper, binds to alpha-7 nicotinic receptors on macrophages and tells them to stop producing TNF. That is the biology behind the 2016 Amsterdam trial, and it is why the therapy is not an analgesic. It is not masking joint pain. It is trying to turn down the production of the molecule driving the disease, using the same nerve that would do it naturally if the feedback loop worked right.
Cut the vagus, inflammation spikes. Stimulate it, TNF drops.
This is the part that makes rheumatologists pay attention even when they are skeptical of implants generally. RA is currently managed with drugs that suppress the immune system systemically: methotrexate, TNF inhibitors, JAK inhibitors, each with its own list of infection risks and monitoring requirements. A device that produces a localized, on-demand version of the same anti-inflammatory signal the body already uses, without circulating drug levels, without a weekly injection, is a genuinely different model of treatment, not just a different delivery method for the same one.
From Borrowed Epilepsy Hardware to a Purpose-Built Implant
The stimulator used in the original Amsterdam trial was adapted from epilepsy technology made by Cyberonics, now part of LivaNova, a device the size of a pocket watch wired into the chest wall. It worked, but it was never designed for this job. SetPoint Medical, a company built directly around Tracey’s research, developed a purpose-built alternative: a microregulator roughly the size of a vitamin capsule that clips around the cervical vagus nerve and is powered and programmed through an external device held near the neck, with no battery or chest incision required. The miniaturization matters because it changes what the surgery looks like, an outpatient procedure under local anesthesia rather than a chest implant, and because it is the version the company has carried into larger controlled trials.
Those trials have not been a straight line to success. A multicenter, randomized, double-blind, sham-controlled study led by Mark Genovese and published in The Lancet Rheumatology in 2020 tested the SetPoint device across sites in the United States and Europe in two stages, an open-label run-in followed by a randomized comparison against sham stimulation. The results were encouraging on safety and on biological plausibility, inflammatory markers moved in the right direction, but the primary efficacy comparison in the randomized stage did not clear the bar with the clarity a drug company would want for a submission. That ambiguity is worth sitting with rather than smoothing over. A therapy with a well-understood mechanism and a strong early pilot still had to go back and run a bigger, harder trial, because seven patients responding to an open-label implant and forty-some patients responding to sham-controlled randomization are different categories of evidence.
What’s New Is Duration, Not Discovery
The news now is not a fresh mechanism. It is an extension study, patients from the controlled trial who continued on active stimulation and were tracked well past the initial endpoint, with results presented to the rheumatology community showing that responders at the original endpoint were, by and large, still responders a year later. Disease activity scores that had come down stayed down. That is a different claim than “it worked once,” and it is the claim that actually matters for a device meant to replace a drug a patient would otherwise take indefinitely.
A year after implantation, the patients who responded were still responding.
Durability is the quiet variable that decides whether a therapy becomes a product or stays a conference poster. Biologic drugs for RA have decades of real-world data on how long a response lasts and how often patients lose response and have to switch agents, a known failure mode called secondary non-response. An implant that has to be surgically placed carries a much higher bar: if the effect fades in eighteen months the way biologic response sometimes does, the cost-benefit math against a drug that can simply be swapped collapses. Twelve months of sustained response is a meaningful data point. It is not yet proof that the effect holds for the five or ten years a patient with a chronic autoimmune disease actually needs.
| Study | Design | Key Finding |
|---|---|---|
| Koopman et al., PNAS, 2016 (Amsterdam/Feinstein Institutes) | Open-label pilot, LivaNova-style cervical VNS implant | Reduced ex vivo TNF production; most patients reached low disease activity or remission |
| Genovese et al., Lancet Rheumatology, 2020 (SetPoint Medical, multicenter) | Two-stage, randomized, double-blind, sham-controlled; purpose-built microregulator | Favorable safety profile; efficacy signal present but not decisive in the randomized stage |
| Extension follow-up (SetPoint Medical cohort) | Open-label continuation of trial responders | Symptom reduction maintained through twelve months, longest reported duration to date |
Who This Actually Helps, and Who It Doesn’t
None of this is being tested as a first-line therapy, and it shouldn’t be. Every cohort in this research program has been drawn from patients who already failed methotrexate and at least one biologic, the population rheumatologists refer to for a reason: standard treatment did not work, and the remaining options get progressively more expensive and immunosuppressive. That is the correct population for a surgical implant to target. Nobody is proposing a vagus nerve stimulator as the first thing offered to a newly diagnosed patient with mild joint pain, and the surgery itself carries real, if low, risks, including hoarseness or vocal cord effects from the stimulation current reaching nearby laryngeal nerve fibers.
The counterargument worth taking seriously is cost and access, not biology. An implant requires a procedure, a device, and a follow-up infrastructure that most rheumatology practices are not built to support, while a biologic drug, however expensive, slots into existing prescribing and insurance workflows that already exist. SetPoint’s device remains investigational in the United States, studied under an FDA investigational device exemption rather than approved for commercial use, and the regulatory and reimbursement path for an implanted neurostimulator competing against a mature biologics market is not short. GSK’s bioelectronics venture, Galvani Bioelectronics, and the broader interest from large pharmaceutical companies in this mechanism suggest the industry believes the biology is real. Believing the biology and building a reimbursed clinical pathway around it are different fights.
The Comparison That Actually Matters
Set this against TNF inhibitors, which is the honest comparison, not against “no treatment.” Biologic drugs work by flooding the bloodstream with an antibody that mops up TNF wherever it finds it, system-wide, continuously, as long as the drug is dosed. A vagus nerve stimulator is trying to restore a regulatory loop that produces the same downstream effect, TNF suppression, but through the nervous system’s own control point rather than through circulating drug. That distinction sounds academic until a patient asks what happens if they stop. Stop a biologic and the drug clears out of the body on its own pharmacokinetic timeline. Stop stimulating the implant and, based on the mechanism as described by Tracey’s group, the suppressive effect should fade on a similarly biological timeline, since nothing about the device permanently reprograms the immune system. That is one of the open questions the twelve-month data does not answer: whether this is a therapy a patient uses continuously for the rest of their life, the same commitment a biologic demands, just delivered through a wire instead of a syringe.
That reframing is the honest way to read this news. The sustained twelve-month data is a real result and a genuine step past the ambiguous 2020 trial. It is not evidence that bioelectronic medicine has leapfrogged biologic drugs for rheumatoid arthritis. It is evidence that the mechanism Kevin Tracey’s lab described in rats twenty-five years ago produces an effect in humans that does not vanish after a few months, which is the minimum bar a device has to clear before anyone should seriously ask an insurer to pay for the surgery that implants it. The question that actually decides whether this becomes standard care is not whether the vagus nerve can be dialed down. It clearly can. The question is whether that dial holds steady for the decade a chronic disease actually requires, and nobody has that answer yet.
Credit: Federico Lancellotti on Unsplash