Transcutaneous Cervical Vagus Nerve Stimulation Improves Speech Comprehension in Noise: A Crossover, Placebo-Controlled Study

Abstract

Background: Speech comprehension in noisy environments remains a significant challenge, even among individuals with clinically normal hearing and users of hearing aids and cochlear implants. Although conventional assistive hearing devices address limitations in the auditory periphery, they do not directly enhance the brain's capacity to segregate speech from background noise. Because tonic vagus nerve stimulation (VNS) has shown potential for rapidly improving central sensory processing, this study investigated whether tonic transcutaneous cervical VNS (tcVNS) can enhance speech-in-noise intelligibility.

Materials and methods: Two cohorts of older human adults (aged 60-84 years) participated in a placebo-controlled, crossover study. Participants completed speech-in-noise assessments using either QuickSIN or AzBio sentences while receiving tonic tcVNS to the neck, or placebo stimulation to the neck-shoulder junction. Speech-in-noise performance was assessed by measuring participants' accuracy in repeating sentences presented at varying signal-to-noise ratios (SNR) within background babble.

Results: Tonic tcVNS improved speech-in-noise intelligibility compared with placebo. At the group level, the SNR threshold for 50% speech intelligibility (SNR-50) improved by 0.76 dB in QuickSIN (p = 0.016) and by 0.38 dB in AzBio (p = 0.045). For individual participants, 50% showed improvements that met a minimum clinically important difference (MCID) of 1 dB. Tonic tcVNS evoked progressively greater improvements as SNR increased in QuickSIN (p = 0.021) and AzBio (p = 0.00023), with the largest gains at SNRs >0 dB. In 55% of participants, tcVNS improved intelligibility beyond an MCID benchmark of 4.9% at 5 dB SNR. Although the magnitude of tcVNS-evoked improvements was inversely related to baseline speech-in-noise impairment (p = 0.028), with the individuals having the most impaired speech-in-noise intelligibility showing the largest gains, it did not correlate with hearing loss severity (p = 0.97) or age (p = 0.88).

Conclusions: Our findings indicate that tonic tcVNS can evoke immediate and clinically meaningful enhancements in speech-in-noise comprehension. This suggests tcVNS may complement conventional assistive hearing technologies and inform novel therapies for sensory processing disorders.

Keywords: Aging; locus coeruleus; norepinephrine; speech in noise; vagus nerve stimulation.

Figure 1.

SIN assessments and stimulation protocol. a. Characteristics of each cohort, along with the study’s primary and secondary outcome measures. b. Timeline of a testing session for the QuickSIN cohort; the AzBio cohort followed an identical timeline. Pure tone audiometry (PTA) was completed without electrical stimulation; data for all participants are shown, with large points and error bars showing the group-average and ±1 SEM connected by a dark black line. Hearing loss cutoffs are shown as different colors in the audiogram. After PTA, tonic tcVNS or off-target stimulation was delivered in a counterbalanced order; the stimulation waveform and frequency are shown beside “tcVNS.” Participants provided an intensity rating of the stimulation after each SIN assessment. [Color figure can be viewed at www.neuromodulationjournal.org]

Figure 2.

Speech intelligibility thresholds improved during tcVNS. SNR-50 estimated with QuickSIN (panel a) and AzBio (panel b). Lower values indicate better SIN performance. Two individuals in the QuickSIN cohort wore hearing aids during the test and are depicted as square markers. Participants in whom tcVNS improvements met or exceeded the MCID are highlighted with darkened outlines. Large data points connected by black lines show group means, with errors ±1 within-subject SEM. p Values are corrected for multiple comparisons using the Benjamini-Hochberg procedure. [Color figure can be viewed at www.neuromodulationjournal.org]

Figure 3.

tcVNS improved speech intelligibility in low-to-moderate noise. Psychometric functions for QuickSIN (panel a) and AzBio (panel b). Smooth lines depict the fitted logistic functions whereas the transparent, jagged lines show the raw accuracy for individual participants. Parameter estimates for SNRα and λ are displayed beside the respective x- and y-axes, with large points connected by black lines indicating the group average and error bars of ±1 within-subject SEM. p Values correspond to the result of linear mixed-effects modeling and are corrected for multiple comparisons using the Benjamini-Hochberg procedure. c. Change in SIN intelligibility as a function of SNR. Positive values indicate better accuracy during tcVNS. Large points and errors bars show the group-average difference and ±1 within-subject SEM; smaller points show individual participants, with those meeting the MCID highlighted with darkened outlines. The best fitting regression lines for each cohort are displayed. d. Percentage change in the odds of accurate SIN comprehension. Positive values indicate higher odds during tcVNS. Points and vertical lines depict the marginal means and their standard error. Star signifiers denote Bonferroni-corrected p-values for post hoc comparisons: ** p < 0.01, *** p < 0.001. [Color figure can be viewed at www.neuromodulationjournal.org]

Figure 4.

Improvements increased with the severity of SIN deficits, but neither hearing loss nor age. tcVNS-evoked changes in the upper asymptote of SIN accuracy (λ) are displayed as function of baseline SNR-50 measured during off-target stimulation (panel a), hearing loss (PTA4) (panel b), and age (panel c). In all panels, each dot depicts individual participants in a cohort; lines depict the marginal regression for each independent variable determined using LME modeling, with shaded areas indicating 95% confidence intervals. All p values were corrected for regression to the mean and for multiple comparisons using the maxT method. [Color figure can be viewed at www.neuromodulationjournal.org]