A Vagus Nerve Pilot Reaches Past Depression, Without a Sham Arm
- Open-label, single-arm pilot, France, N = 17 adults with long COVID (94% women, mean age 49.8 ± 6.6).
- taVNS protocol: left ear (cymba conchae), 2 Hz, 200-microsecond pulse width, 1 to 8 mA individually titrated, 1 hour per session, once weekly for 8 weeks.
- Autonomic score (COMPASS-31): 33.71 to 13.78, a 56.6% drop, p < 0.0001, 95% CI [-25.47, -13.97] – survived correction for testing three outcomes at once.
- PTSD checklist (PCL-5): 25.76 to 19.47, a 24.4% drop, p = 0.028, 95% CI [-13.18, -0.82] – did not survive that same correction (FDR p = 0.07).
Vagus nerve stimulation earned its first psychiatric approval in an implanted, surgical form, for treatment-resistant depression – a specific device for a specific diagnosis. The pilot study below tests a different device, worn rather than implanted, on a population where depression is not the entry criterion at all. Long COVID brought the patients in; only one of three measured outcomes touches a psychiatric symptom, and that one carries a footnote the other two do not.
A Different Device, a Different Population
The therapy tested here is transcutaneous auricular vagus nerve stimulation (taVNS): a clip placed on the cymba conchae of the left ear, not a surgically implanted pulse generator. Seventeen adults with long COVID, 94% of them women, mean age 49.8, were recruited at a rehabilitation department in the Paris region between 2023 and 2024. All seventeen had lived with long COVID symptoms for at least 18 months, some for as long as 4.5 years, and all scored below 50 on a standard quality-of-life scale at entry. Every participant received the same treatment: a TENS Eco+ device delivering a 2 Hz signal at 200-microsecond pulse width, current individually set between 1 and 8 milliamps, once a week for an hour, across eight consecutive weeks. There was no comparison group. Each person's outcome was measured against their own baseline.
One Signal Among Three, and Its Fine Print
Three outcomes were specified in advance: a dysautonomia score (COMPASS-31), a cognitive test battery (COGBAT), and a PTSD symptom checklist (PCL-5). The autonomic score dropped 56.6%, from 33.71 to 13.78 (p < 0.0001), and stayed significant after the authors corrected for testing three outcomes at once. The cognitive score rose 31.0% (p = 0.011), also surviving correction. The PTSD checklist fell 24.4%, from 25.76 to 19.47 (p = 0.028); once the same correction was applied, though, that result no longer cleared the bar (FDR p = 0.07). The authors describe it themselves as an exploratory signal, not a confirmed effect, and note that an open-label design without a sham arm carries a particular risk of expectation bias on self-report scales like PCL-5 and COMPASS-31. No adverse events were reported.
The autonomic-symptom score cleared statistical correction. The PTSD-symptom score, the study's only psychiatric measure, did not.
Open-label, single-arm pilot with 17 participants and no sham comparator; the authors themselves name expectation and effort-to-please bias as a specific risk for self-report measures including the PTSD checklist; no formal sample-size calculation, only a feasibility threshold of 15 participants; no follow-up after the eight-week course ended; the sample is 94% female, which the authors flag as a limit on generalizing to men; the PTSD result specifically did not survive correction for multiple comparisons.