A systematic review of 12 randomized trial publications involving 908 people with tinnitus and anxiety found higher reported response rates and lower scores on several tinnitus and anxiety measures in acupuncture-based treatment groups than in control groups. The strength of the evidence varied considerably across outcomes.
Across nine studies, 329 of 368 participants in the treatment groups were recorded as having a response, or 89.4%, compared with 281 of 371 people in control groups, or 75.7%. The pooled relative risk was 1.18, with a 95% confidence interval from 1.10 to 1.26 and a P value below 0.00001. The analysis reported I-squared of 0%, a statistic used to show how much results differ between studies.
Symptom scores generally favored acupuncture groups
Tinnitus-specific measures also pointed in the same direction. In seven trials, scores on the Tinnitus Severity Rating Scale, known as the TEQ, were lower in treatment groups by a pooled mean difference of 1.96 points. The 95% confidence interval ranged from 1.45 to 2.47 points lower, with a P value below 0.00001 and I-squared of 48%.
In nine trials using the Tinnitus Handicap Inventory, or THI, scores were 9.11 points lower on average in treatment groups. The 95% confidence interval ranged from 5.34 to 12.89 points lower, and the P value was below 0.00001. However, I-squared was 95%, indicating very high variation among the study results.
The anxiety findings followed a similar pattern, although each analysis drew on only a small part of the review. Four trials using the Hamilton Anxiety Rating Scale found scores 2.04 points lower in treatment groups, with I-squared of 82%. Two trials using the Self-Rating Anxiety Scale found scores 3.50 points lower, with I-squared of 78%. Two trials using the Beck Anxiety Inventory found scores 8.80 points lower, with I-squared of 66%. Each pooled comparison had a P value below 0.05.
Only four trials reported traditional Chinese medicine syndrome scores. The pooled difference favored treatment groups by 0.59 score units, but I-squared was 96%. After one study was removed in a sensitivity analysis, the reported heterogeneity fell to zero, with P equal to 0.67 and I-squared equal to 0%.
Biomarker results were inconclusive
The biological measures did not provide the same clear signal. Only two trials contributed data for each of BDNF, 5-HT and GABA, and none of the pooled comparisons showed a statistically significant difference between groups. For BDNF, the mean difference was minus 72.27, with a 95% confidence interval from minus 312.41 to 167.87, P equal to 0.56 and I-squared of 80%.
For 5-HT, the mean difference was minus 22.56, with a 95% confidence interval from minus 56.12 to 10.99, P equal to 0.19 and I-squared of 99%. For GABA, the mean difference was 56.23, with a 95% confidence interval from minus 46.50 to 158.96, P equal to 0.28 and I-squared of 97%. The high heterogeneity and small number of contributing trials limit what these comparisons can show.
The evidence came from a varied set of trials
Eligible studies were Chinese- or English-language randomized controlled trials involving people whose main symptoms were tinnitus and anxiety. Conventional acupuncture could be used alone or alongside Chinese or Western therapy, while control groups included sham acupuncture or Western medication. The review searched nine databases from their starting dates through January 15, 2026. The initial search produced 949 records; after removing 288 duplicates, excluding 434 records by title or abstract and rejecting 215 full texts, 12 documents remained.
Two researchers independently screened and evaluated the studies, cross-checked their data extraction and used a third researcher to resolve disagreements. They contacted original authors when data were missing and rejected studies when missing information could not be recovered. Risk of bias was assessed with the RoB2 tool across five domains. The review was registered in PROSPERO as CRD42024596594 and stated that it would follow PRISMA-P.
The analysis used RevMan 5.3. It combined continuous scores with weighted mean differences and the categorical overall response outcome with relative risk, reporting 95% confidence intervals. P values below 0.05 were treated as statistically significant, while I-squared and heterogeneity tests helped determine whether fixed-effects or random-effects models were used.
The review did not treat every finding as equally reliable. GRADE rated three evidence items as high quality, one as moderate and six as low quality, with downgrading mainly because of risk of bias, unexplained heterogeneity and imprecision. The authors also identified possible publication bias, incomplete outcome measurement, and a lack of safety and long-term efficacy assessment. Differences in acupuncture protocols, interventions, treatment lengths and efficacy criteria added further uncertainty.
For the overall response outcome, a visually symmetric funnel plot suggested a low probability of publication bias, but the review described that assessment as subjective and potentially biased. The authors reported support from the Master of Traditional Chinese Medicine, National Famous Chinese Medicine Practitioner, Qihuang Scholar Anhui Province Seed Player Training Project and the 2025 Anhui Provincial Department of Education Project.
Taken together, the pooled results favor acupuncture-based treatment strategies for reported response rates and symptom scores in people with tinnitus and anxiety. They do not settle whether acupuncture changes BDNF, 5-HT or GABA, establish long-term efficacy or safety, or identify one best acupuncture protocol. The review leaves those questions open because the evidence combined varied treatments and comparators, and several outcomes were based on only two to four trials.
Paper data and sources
Original title: Efficacy of acupuncture for tinnitus with anxiety: a systematic review and meta-analysis
Authors: Peiyun W, Shuaihui Q, Yi C, Hao Y
Journal/Repository: Frontiers in neurology
Status: Peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text