How to read a vagus nerve stimulation study

Begin with the research question, not the abstract conclusion. A mechanistic study can show that a pathway or marker changed. A feasibility study can show that a procedure was workable. A safety study can describe adverse events at a tested exposure. A randomized trial can estimate a between-group effect. Those answers are useful, but they are not interchangeable.

The label taVNS is not a complete intervention description. Readers need the participant group, exact device, electrode, anatomical site, side of stimulation, waveform and total exposure. Farmer and colleagues proposed tVNS-specific minimum reporting standards because incomplete method descriptions prevent replication and responsible comparison.

1. Define the population: age, health or diagnosis, baseline severity, setting, exclusions and sample size.
2. Identify the exact hardware and software: manufacturer, model, electrode, coupling medium, channel and control interface.
3. Locate stimulation precisely: cervical or auricular, left or right, named ear region, active and reference electrode positions.
4. Reconstruct dose: current or intensity rule, frequency, pulse width, waveform, duty cycle, ramping, session duration, number of sessions and follow-up.
5. Check what was actually delivered, not only what the protocol planned, including adherence, modifications and fidelity.

A controlled study estimates a difference between conditions. The interpretation therefore depends on what the comparison group experienced. In electrical stimulation research, sham may imitate contact, sound, attention or sensation, but it can also produce somatic or physiological effects. ‘Sham-controlled’ is not enough detail by itself.

A p value addresses compatibility with a statistical model under a null hypothesis. It does not tell you whether an effect is large, useful, precise, unbiased or relevant to a new product. Read the estimate, interval, units, baseline risk and outcome scale together.

Target engagement asks whether an intervention affected its predicted target. A biomarker or imaging response may support that question. A clinical or lived-experience outcome asks whether participants improved in a meaningful way. One layer can strengthen a mechanistic argument without proving the next layer.

Risk of bias is not a general impression of whether a paper looks scientific. It asks whether features of the design, conduct, analysis or reporting could systematically move the result away from the truth. RoB 2 organizes this for randomized trials across randomisation, deviations from intended interventions, missing outcomes, outcome measurement and selection of the reported result.

1. Compare the trial registry, protocol or analysis plan with the published outcomes and analyses.
2. Check whether allocation was concealed and whether baseline differences suggest a problem with randomisation.
3. Identify deviations from the assigned intervention and whether analyses matched the effect the study intended to estimate.
4. Count missing outcomes and compare dropout reasons between groups; do not rely only on the percentage who completed.
5. Ask whether outcome assessors could be influenced by knowledge of allocation or by participant expectations.
6. Check whether harms were actively solicited, defined, timed, graded and reported with group denominators and discontinuations.
7. Read funding, device supply, author conflicts, data access and the role of the sponsor without assuming that disclosure alone resolves bias.

A systematic review uses explicit methods to identify, select and appraise relevant studies. A meta-analysis statistically combines compatible effect estimates. A review may use narrative synthesis instead when studies are too different. Neither label guarantees that the underlying evidence is complete, unbiased or directly applicable.

The last step is not ‘Was the study positive?’ It is ‘What exactly can this result support?’ Applicability depends on the match between participants, intervention, comparator, outcome, setting and follow-up. Product transfer adds another layer: hardware, placement, protocol, intended use and regulatory context should be relevant to the new device and claim.

• A finding from implanted VNS does not automatically support cervical or auricular non-invasive stimulation.
• A finding from one taVNS device or ear site does not automatically support another electrode, waveform or dose.
• A physiological or imaging response does not automatically support a symptom or wellness claim.
• A clinical population and indication do not automatically support general wellness use in healthy consumers.
• A field-level review can map the category while still providing no product-specific evidence for Neuvago.
• A negative or inconclusive study can be informative about the target, protocol, comparator, measurement or limits of the hypothesis.