Chronic subjective tinnitus is associated with maladaptive activity across auditory and non-auditory networks. Therapeutic effects of repetitive transcranial magnetic stimulation (rTMS) to modulate brain activity are uneven due to diversity in stimulation protocols, patient characteristics, outcome assessments, and control conditions. To systematically review rTMS clinical data for persistent subjective tinnitus and examine how stimulation frequency, cortical target, dose, localization, comparator design, and participant characteristics influence treatment response. PRISMA 2020 was used for the systematic review using PICOTS eligibility criteria. We searched PubMed/MEDLINE, Embase, the Cochrane Library, and Web of Science using tinnitus and rTMS-related controlled vocabulary and free-text keywords. A multi-author sequential selection process examined titles, abstracts, and full texts. Participant data, trial design, stimulation target and parameters, localization, comparison, follow-up, tinnitus outcomes, responder criteria, adverse events, and primary results were extracted. Risk-of-Bias was assessed using validated design-specific tools, including RoB 2 for randomized trials and ROBINS-I for non-randomized comparative studies. Because of substantial clinical and procedural heterogeneity, findings were synthesized qualitatively rather than statistically pooled. Twenty-eight studies were included in the primary qualitative synthesis; low-frequency 1-Hz auditory or temporoparietal rTMS was most common. Tinnitus burden decreased in small and early investigations, but several larger sham-controlled trials did not demonstrate consistent superiority over sham. High-frequency temporal stimulation was associated with transient reductions in tinnitus perception rather than durable therapeutic improvement. Combined auditory–prefrontal or multisite protocols showed more sustained improvement in selected studies, although findings were inconsistent. THI, TQ, and perceptual ratings were analyzed separately because they represent distinct tinnitus outcome domains. Stimulation dose, target location, sham credibility, follow-up, hearing condition, tinnitus phenotype, and responder criteria varied greatly. In safety trials, rTMS was generally feasible and well-tolerated; however, transient worsening of tinnitus and treatment-related withdrawal occurred in selected studies, and safety reporting was inconsistent. No specific rTMS protocol is currently supported for routine clinical treatment of chronic subjective tinnitus. Combined auditory-prefrontal and multisite procedures are helpful in certain trials, especially for reducing both perceptual and tinnitus-related strain but have not consistently surpassed single-site stimulation.
Browser-based hearing tests offer a scalable alternative to clinic-based assessment, but most validated tools have been developed in English-speaking populations. The authors conducted a retrospective observational analysis of real-world use of a Japanese browser-based hearing test platform (April 2023–March 2026), complemented by a prospective multicentre paired sub-study against conventional clinical speech audiometry. The retrospective analysis included 8,022 unique users and 86,959 test executions across four modules: music perception (Montreal Battery of Evaluation of Musical Abilities); speech perception in quiet; speech in noise at +10, +5, and 0 dB signal-to-noise ratios (SNRs); and consonant identification. Completion rates of the full-length modules ranged from 43.7% to 85.8%. Speech perception in quiet showed pronounced upper-end clustering (41.0% of completed attempts and 57.9% of users scored ≥90%), whereas music perception was more broadly distributed (20.8% and 21.2%, ≥90%), and speech-in-noise accuracy fell progressively from a mean of 72.0% at +10 dB to 57.0% at 0 dB SNR, where no attempt reached 90%. These patterns persisted when the analysis was restricted to one attempt per user. Music perception correlated only modestly with all speech-related outcomes (Spearman ρ = 0.31–0.39). Repeat administration by the same user yielded intraclass correlation coefficients of 0.58–0.85 and a smallest detectable change of 17.6–35.8 percentage points. In the sub-study, 41 participants (25 unaided, 16 aided) completed the conventional 20-item 67-S word recognition test in a calibrated booth and, within days to weeks, the 20-item online check unsupervised on their own device. Agreement was moderate (Pearson r = 0.65, 95% CI [0.43, 0.80]; ICC(2,1) = 0.62, 95% CI [0.37, 0.78]), with a small negative bias (−6.2 percentage points, 95% CI [−10.8, −1.6]) and wide limits of agreement (−34.7 to +22.2). These data characterize the concordance of a non-English digital hearing assessment platform with clinical speech audiometry and provide the first estimates of its measurement error. They do not establish screening accuracy: no pure-tone reference, referral threshold, sensitivity, or specificity was evaluated, and the limits of agreement are too wide for individual-level clinical interpretation.
Two small public corpora, the PhysioNet earndb tone-pip set and the Mendeley click set, recur as evaluation data for objective auditory brainstem response (ABR) detection. Neither corpus includes expert annotations of response presence. Studies that cannot commission their own annotations therefore construct the label automatically, most commonly by declaring a response present when two repeated sub-averages correlate above a threshold—the clinical two-replication reproducibility criterion. Whether detector comparisons built on that automatic label would survive a switch to expert annotation has not been tested. Audiologist annotations published by Liu et al. (2025) were mapped onto the recordings (one rater, 204 earndb conditions; six raters, 86 Mendeley conditions) and the mapping validated against the record inventories. On identical data, we constructed the reproducibility surrogate and four detectors spanning the capacity range (the instrument statistic wnsfmp, logistic regression, gradient boosting, a multilayer perceptron). Each learned detector was trained under each label and scored against each label (a 2 × 2 factorial) by leave-one-subject-out cross-validation; discrimination is reported pooled and as a frequency-adjusted AUROC; inference rests on exact subject-level permutation tests. The surrogate agreed with expert judgment at κ = 0.167 (earndb) and 0.248 (Mendeley), against κ = 0.691 among the six Mendeley audiologists. It is largely a function of signal-to-noise ratio: a single-replicate instrument statistic predicted it with AUROC 0.789 on sub-averages in which the expert saw no response, and a simulation with no response reproduced this. Training on the surrogate cost 0.139–0.304 pooled AUROC against expert judgment (perceptron: every informative listener, exact p = 0.031), a loss arising across listeners and frequencies rather than within them; scoring fixed predictions against either label did not reorder detectors. Changing both, as a study would, reordered them: wnsfmp fell from 0.827 to 0.662, while the perceptron rose from 0.713 to 0.898 (interaction −0.350, exact p = 0.031); about 37% of this is pooling across frequencies with different expert prevalence, and the within-frequency remainder was not resolved. Rescoring against individual Mendeley raters moved AUROC by 0.065–0.082, below the smallest difference eight listeners can resolve (0.096–0.136). The stimulus level alone reached AUROC 0.931 within 1 kHz, exceeding every detector there. On these corpora, measured detector performance depends on how the label is constructed and who provided it, at magnitudes equal to or larger than the differences between detectors. The label acts chiefly through what detectors learn, and the proxy leaks the input. We claim neither label is correct; we show that the choice determines the answer and release the mappings and a CPU-reproducible pipeline.
Marc-Olivier Mainville, Émilie Lacasse-Rioux, Arnaud J. Noreña, Guillaume T. Vallet, Philippe Fournier
Tinnitus, post-traumatic stress disorder (PTSD), and chronic pain are prevalent chronic conditions among veterans and frequently co-occur. These conditions have been associated with alterations in autonomic nervous system (ANS) functioning. Heart rate variability biofeedback (HRVB) is a non-pharmacological intervention intended to modulate autonomic regulation. This scoping review aimed to identify HRVB interventions used for tinnitus, PTSD, and chronic pain in veterans and evaluate the extent and characteristics of the existing evidence in this population. A systematic search of PubMed, Embase, and CINAHL was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews. Studies examining the effects of HRVB, alone or in combination with other non-pharmacological interventions, on chronic symptoms in veteran-only populations were included. Of 69 records identified, five studies met inclusion criteria. Four studies examined HRVB in relation to PTSD symptoms, with mixed findings: two reported improvements in overall PTSD symptoms, while one reported reduction limited to intrusion symptoms. One study also reported improvements in depressive symptoms. A single study examined chronic pain outcomes and reported reductions in perceived pain following HRVB training. No studies investigated HRVB in relation to tinnitus. Across studies, small sample sizes, methodological heterogeneity, and variability in HRV measurement approaches limited comparability and interpretation of findings. Current evidence on HRVB in veterans is limited and does not allow conclusions regarding its efficacy as a standalone intervention for PTSD or chronic pain, although it may have potential as an adjunctive or supportive approach. Notably, no studies have examined HRVB in relation to tinnitus in veterans. These findings highlight the need for rigorously designed studies to evaluate the feasibility, safety, and potential relevance of HRVB in well-characterized veteran populations, including those with tinnitus and co-occurring conditions.
