Stellate ganglion block has not been shown to help tinnitus enough to notice: a 2026 meta-analysis of 11 trials found a 5.73-point handicap improvement, below every published threshold for noticeable change, testing the block only in combination.
Most of the treatments on this site aim at the brain or the ear. This one aims at the nerves that control blood vessels — and it involves a needle in the neck.
A stellate ganglion block is an injection of local anaesthetic near the cervical sympathetic chain, a cluster of nerves low in the neck that regulates blood flow, pupil size and sweating on that side of the head. It is an established procedure used mainly for certain pain conditions, and it has been tried for tinnitus on a straightforward rationale: if the problem involves inadequate blood supply to the inner ear, or a nervous system stuck in a state of arousal, interrupting sympathetic signalling might address either.
In 2026 that idea got its first proper pooled evaluation, and the results look excellent until you read them closely.
What the meta-analysis reports
A systematic review and meta-analysis in Frontiers in Neurology pooled 11 randomised controlled trials covering 915 patients. What it found:
- Overall effective rate — odds ratio 4.53 (95% CI 3.15–6.53), p < 0.00001
- Tinnitus Handicap Inventory — mean difference −5.73 (−6.10 to −5.36), p < 0.00001
- Self-Rating Anxiety Scale — mean difference −11.37 (−12.46 to −10.29), p < 0.00001
- Basilar artery flow velocity — systolic +5.60 cm/s (4.40–6.80), diastolic +4.26 cm/s (3.70–4.83)
- Carotid artery flow velocity — peak systolic +4.73 cm/s (3.26–6.18), end-diastolic +10.85 cm/s (6.02–15.68)
Every one of those is statistically significant, most of them overwhelmingly so. An odds ratio of 4.53 would put this among the most effective things in the field. That is exactly why the rest of this article exists.
The number that matters most is below the line
Start with the Tinnitus Handicap Inventory, because it is the only outcome here that directly measures how much tinnitus is affecting a person's life.
The THI runs from 0 to 100. The pooled improvement was 5.73 points.
For scale, use a threshold this site already relies on elsewhere: the TENT-A3 pivotal trial of a bimodal stimulation device defined a responder as someone improving by at least 7 points on the THI. That was a definition chosen for one trial rather than a measurement of what patients notice — but no available estimate of a noticeable change is lower.
The pooled effect of stellate ganglion block lands beneath it.
It lands further beneath the other available estimate. A 2025 multi-centre study anchoring the threshold to patients' own reports of improvement put the minimal clinically important difference on the THI at 11 points, with a range of 7.8 to 12. Against 7 or against 11, 5.73 is below every published figure.
That is the single most important sentence in this article, and it survives taking the meta-analysis completely at face value. A p-value below 0.00001 tells you an effect is unlikely to be zero. It tells you nothing about whether the effect is big enough to notice. With 915 patients, a change far too small to feel can still be measured with enormous statistical confidence — and here it was.
The confidence intervals do not behave like tinnitus data
Now look at how that −5.73 is reported: 95% CI −6.10 to −5.36. The entire interval is less than three-quarters of a point wide.
That interval is claiming the pooled effect is known to within a fraction of a THI point, on the basis of 11 trials run by different teams, on different patients, with different protocols, measuring a subjective symptom by questionnaire.
Tinnitus does not produce data like that. Compare it with the other pooled estimates on this site: tDCS reports a standardised mean difference of −0.35 for loudness with an interval spanning a substantial range, and the whole field is characterised by heterogeneity so consistent that reviews comment on it as a matter of routine. The anxiety-scale interval here has the same signature — −11.37, from −12.46 to −10.29, again barely two points wide.
Confidence intervals this narrow mean the included trials agreed with one another almost perfectly. In a field where trials routinely disagree, near-perfect agreement is not reassurance. It is the same warning sign the 2025 ginkgo network meta-analysis raised from the opposite direction, where an implausibly large effect pointed at the trials rather than the treatment. When pooled results look better behaved than the field they come from, the explanation is usually in the trials.
The review's authors are not claiming otherwise. Their own conclusion calls for future large-scale rigorous trials to standardise treatment, address potential bias, and confirm long-term benefit.
It tested the combination, not the block
There is a structural limit here that no amount of statistical care gets around.
What the meta-analysis compared was the block combined with another therapy against a control group. That is a test of a package, not of an ingredient. If the combination outperforms the control, the benefit could belong to the block, to whatever it was added to, or to something about receiving a procedure at all — and pooling more trials of the same design cannot separate them.
This matters more than usual for an invasive treatment. The question a patient faces is not "does this package help" but "is the needle worth it", and that is precisely the comparison the evidence does not make.
Blood flow is a mechanism, not a benefit
The haemodynamic findings are the most solid part of the paper and the least relevant to a patient.
Increasing basilar artery systolic velocity by 5.60 cm/s demonstrates that the block did what a sympathetic block is supposed to do. It confirms the procedure worked pharmacologically. It is genuine evidence for the mechanism.
It is not evidence that anyone's tinnitus got better. A measurement showing a drug reached its target is not a measurement of benefit, and these two things are easy to conflate when they appear in the same results table under the same p-values. The outcome that speaks to benefit is the THI — and that one came in below the responder threshold.
The anaesthetic question the abstract leaves open
Here is a confound specific to this treatment, and it cuts at the mechanism directly.
Intravenous lidocaine is documented to suppress tinnitus temporarily in a minority of people. It is one of the most interesting findings in tinnitus pharmacology precisely because the effect is real — and unusable, since it requires infusion and carries cardiac risks that cannot be justified for a non-life-threatening symptom.
A stellate ganglion block delivers local anaesthetic into a well-vascularised part of the neck. If any meaningful quantity reaches the systemic circulation, then short-term tinnitus suppression after the procedure could reflect the anaesthetic doing what local anaesthetics are already known to do — rather than the sympathetic block achieving anything.
Separating those requires a control designed for it: the same drug delivered somewhere it cannot block the ganglion. Whether any of the 11 trials did that is not something the abstract answers, and it is the first question worth asking of the full text.
The anxiety score is the largest change reported
One finding deserves to be taken seriously rather than explained away. The Self-Rating Anxiety Scale moved by 11.37 points — proportionally the biggest effect in the paper, and considerably larger than the tinnitus handicap change.
That is a coherent result. Blocking the sympathetic chain is, fairly directly, an intervention on the body's arousal machinery, and stellate ganglion block is being studied for anxiety-related conditions in its own right.
But note what it implies for this question. If the intervention moves anxiety substantially and tinnitus handicap sub-threshold, the most natural reading is that what improved was arousal, with tinnitus benefiting slightly and indirectly. That is a real route to feeling better — it is roughly the route CBT takes — but CBT reaches it without a needle, with a far larger evidence base, and with durable effects rather than the duration of a local anaesthetic.
What the procedure involves
Anaesthetic is injected into the neck near the sympathetic chain, usually with ultrasound or X-ray guidance to place the needle among a crowded set of structures.
A successful block typically produces a drooping eyelid and constricted pupil on that side — Horner's syndrome — which resolves as the anaesthetic wears off. That is the confirmation the block worked, not a complication. Hoarseness and a numb-feeling throat are common and similarly temporary.
The serious risks are uncommon and come from the anatomy the needle passes near: injection into a blood vessel, seizure, collapsed lung, and spread of anaesthetic to spinal structures. The meta-analysis is titled as covering efficacy and safety; the outcomes it reports are efficacy and blood flow, and this site cannot quote a pooled complication rate from it.
Where this leaves it
None of this makes stellate ganglion block worthless: it is a real procedure with a real mechanism, tested in real randomised trials. What the evidence leaves it is unproven in the specific way that matters: nobody yet knows what the block on its own does to tinnitus in a patient who would feel the difference.
If you have exhausted the standard options and are looking at this, the useful comparison is not against nothing. It is against the treatments with the strongest evidence, which are non-invasive, carry no procedural risk, and are judged by the same yardsticks this one falls short of: the 7-point responder cut-off and the 11-point estimate of a change patients notice. That comparison is the conversation to have with a specialist who knows your case — and when treatments haven't worked covers what else is on the table before a needle is.