The vagus nerve is the body's longest cranial nerve, a two-way cable running from the brainstem down through the neck to the chest and gut, and most of its traffic travels upward, from body to brain. That anatomy has made it irresistible to anyone looking for a back door into failing memory circuits: clip an electrode to the outer ear, where a branch of the vagus runs just under the skin, and you are, in theory, sending a signal into the brainstem without a scalpel. Devices that do exactly this are now cheap and widely available. So a group of clinicians at Sheffield Teaching Hospitals NHS Foundation Trust and the University of Sheffield went looking for the evidence that any of it helps people with memory loss. They searched PubMed, Cochrane, MEDLINE, EMBASE and Google Scholar between 1 August 2025 and 1 April 2026, and checked the clinical trials register on 10 February 2026. They came back with six studies.
Six. That is the whole human interventional literature for vagus nerve stimulation specifically in mild cognitive impairment (MCI) or Alzheimer's disease, according to the review by Ruth Narramore and colleagues, published in NeuroSci on 27 April 2026. One used an implanted stimulator (iVNS); five used transcutaneous stimulation at the ear (taVNS). Five of the six enrolled 60 participants or fewer. The proposed biology is genuinely interesting: vagal signals are thought to reach the locus coeruleus and the noradrenaline system, and the cholinergic pathways that degenerate early in Alzheimer's. But the authors are blunt that these mechanisms are largely derived from animal models and have yet to be proven in human clinical studies. A pathway that works in a rodent is a hypothesis about a person, not a finding in one.
The oldest and strangest entry is the implant study. In 2006 a Swedish team reported one year of follow-up in 17 people with Alzheimer's who had a vagus nerve stimulator surgically implanted, a non-randomised, self-controlled design with single blinding. The review's tables lay out what happened: at six months, 13 of 17 showed no decline or improvement on the ADAS-Cog, a median change of three points that reached statistical significance; by one year that had fallen to 7 of 17, with a median change of minus two points that did not. On the MMSE, 12 of 17 were stable or better at both six months and a year. Cerebrospinal fluid was sampled at baseline, three months and twelve months: tau levels varied enormously at baseline (99 to 764 pg/mL) and showed no statistically significant change, and there was no difference in beta amyloid or Abeta42 either. Hoarseness and cough were the most common side effects, mild to moderate. That is the strongest single-patient-level dataset in this field, and it is a 17-person uncontrolled pilot from two decades ago.
The non-invasive studies are newer and friendlier, and mostly not designed to answer the question families actually ask. The review notes that most were built around safety, feasibility and mechanistic biomarkers rather than treatment effect, with one exception: Wang and colleagues powered their double-blinded randomised trial to detect a difference on the MoCA-B, randomising 60 people with MCI to ear stimulation or sham for 24 weeks of home self-administration, with 52 analysed and a significant difference in overall MoCA-B scores. Across the six studies the signal, such as it is, sits in working memory domains including verbal fluency and three-dimensional processing, with global cognitive scores sensitive in some cases. Stimulation settings, electrode sites and schedules differ from study to study, and the reviewers make a quietly damning observation about how those settings were chosen: "Justification for protocols seemed to lean towards practicality over clinical efficacy." Nobody yet knows the right dose of this, or whether dose is even the right concept.
What the review does establish is worth something. Non-invasive vagus stimulation was well tolerated in this population, and people with MCI and mild Alzheimer's were able to apply the devices themselves, at home, over prolonged periods. That is not nothing: an intervention that older adults with memory problems can actually carry out unsupervised clears a hurdle that defeats many trial designs. The authors treat this as groundwork, evidence that properly blinded, adequately powered, long-term studies are feasible rather than evidence that the treatment works. They also flag a gap worth watching: there are no human VNS studies pairing stimulation with another intervention, though pairing has improved results in stroke work, and one planned study would combine ear stimulation with transcranial magnetic stimulation.
For a family reading the marketing copy on an ear-clip device, the honest translation is this. The heterogeneity of the six studies is so great that the reviewers say no robust pooled estimate of efficacy is possible. Diagnostic approaches varied, with limited use of Alzheimer's-specific biomarkers or advanced imaging, so the findings may not generalise to people whose Alzheimer's has been biomarker-confirmed. Tolerable and feasible are real findings. Effective is a different word, and the evidence base has not earned it yet. Anyone weighing a device, a trial, or a change to what they are already doing should be having that conversation with the clinician who knows their diagnosis.