Complex auditory brainstem responses and speech-in-noise perception in young workers with occupational noise exposure

1 Introduction

The World Health Organization (WHO), in its 2021 World Report on Hearing, estimates that unsafe listening practices, such as exposure to loud noise in occupational settings and social environments, place half of the global population at risk of developing hearing loss (Organization WH, 2021). In addition, numerous experimental studies in animals have demonstrated that moderate to high levels of noise exposure can cause permanent damage to synapses between inner hair cells and auditory nerve fibers, even in the absence of outer hair cell (OHC) damage and clinically significant permanent threshold shifts (Kujawa and Liberman, 2009; Lin et al., 2011; Valero et al., 2017).

This synaptic damage leads to a progressive loss of auditory nerve fibers, though not all fibers are affected equally. Animal studies suggest a preferential loss of low spontaneous rate auditory fibers (Furman et al., 2013), which are not essential for encoding low-level sounds and, therefore, do not contribute to elevated hearing thresholds (Bourien et al., 2014). The degeneration of these synapses, a condition known as cochlear synaptopathy, is thought to be a key mechanism underlying hidden hearing loss (HHL) (Valderrama et al., 2022). While cochlear synaptopathy has been strongly linked to noise exposure, it has also been observed as part of the normal aging process, even in individuals without a history of excessive noise exposure (Kujawa and Liberman, 2009; Sergeyenko et al., 2013).

Because high spontaneous rate fibers remain intact, cochlear synaptopathy can occur despite normal audiometric thresholds and preserved OHC function, as assessed with otoacoustic emissions (Kujawa and Liberman, 2009; Furman et al., 2013). However, a major challenge in detecting cochlear synaptopathy in humans is the lack of non-invasive diagnostic tools, as the methods used in animal research (e.g., immunostaining) are not clinically viable. Most human studies have explored potential diagnostic markers using behavioral and electrophysiological measures, particularly suprathreshold auditory brainstem responses (ABR) and speech perception in noise tests.

The loss of low spontaneous rate/high-threshold auditory nerve fibers may lead to a reduction in the fidelity of sound encoding at medium-to-high intensities. This deficit is particularly detrimental in low signal-to-noise ratio (SNR) conditions, where it can impair fine-structure and envelope coding (Kujawa and Liberman, 2009; Bharadwaj et al., 2014). Thus, given that cochlear synaptopathy disrupts the neural encoding of sound stimuli, the frequency-following response (FFR) is likely to be a potential tool for detecting this condition in humans.

The FFR, measured via complex auditory brainstem responses (c-ABR), is an electrophysiological marker of sustained neural temporal coding, reflecting phase-locked neural activity (Plack et al., 2014). Animal studies support this approach. Two studies demonstrated that FFR strength was reduced in mice with histologically confirmed cochlear synaptopathy, whether induced by aging (Parthasarathy and Kujawa, 2018) or noise exposure (Shaheen et al., 2015).

In humans with a history of noise exposure, studies have reported alterations in FFR metrics, including changes in the FFR envelope slope (Bharadwaj et al., 2015), reductions in fundamental frequency (F0) magnitude (Suresh and Krishnan, 2022), and decreases in overall FFR magnitude (Bramhall et al., 2021). Additionally, FFR has been shown to predict self-reported difficulties in speech perception in noise (Anderson et al., 2013).

However, findings from studies investigating FFR as a biomarker of cochlear synaptopathy in young adults with normal hearing remain inconsistent. While some studies have found associations between FFR outcomes, speech-in-noise perception, and noise exposure history, others have failed to replicate these findings (Suresh and Krishnan, 2022; Ananthakrishnan et al., 2023; Grose et al., 2017; Guest et al., 2018; Prendergast et al., 2017). These discrepancies highlight the need for further research to refine FFR-based diagnostic tools and to better understand the variability in individual susceptibility to cochlear synaptopathy.

Most previous research on noise-induced cochlear synaptopathy has focused on young adults with recreational noise exposure or firearm use, while relatively little is known about the effects of occupational noise exposure. If cochlear synaptopathy occurs in humans, workers exposed to noise may be at risk of developing this condition before exhibiting permanent threshold shifts.

A proposed perceptual consequence of cochlear synaptopathy is difficulty understanding speech in complex acoustic environments, even when audiometric thresholds remain within normal limits (Liberman et al., 2016). Therefore, we hypothesize that noise-exposed workers may develop cochlear synaptopathy, which could be detected through complex auditory brainstem responses (c-ABR) and speech-in-noise perception tasks.

Given the potentially high prevalence of HHL, its risk of progression to more severe auditory impairment, and the absence of standardized clinical diagnostic protocols, there is an urgent need to improve HHL detection (Valderrama et al., 2022). Advancing diagnostic methods may facilitate the development of prevention programs aimed at mitigating the progression of hearing impairment in noise-exposed populations.

The objective of our study is to contribute to this effort by investigating whether c-ABR outcomes and speech-in-noise perception performance are associated with occupational noise exposure in young workers with normal hearing thresholds and preserved otoacoustic emissions.

2 Methods

2.1 Participants

Two groups of participants, both aged between 18 and 40 years, were selected from Zhejiang province, China. One group consisted of 44 participants exposed to occupational noise, while the other comprised 44 participants not exposed to occupational noise serving as controls. Participants were classified into the noise-exposed group if they had been exposed to noise levels of 80 dBA or higher (based on the normalized equivalent continuous A-weighted sound pressure level over 8 h, LAeq, 8h) for more than 2 years in the same industry. Single noise measurements were taken at their workplaces. The control group included participants from industries or workplaces with noise levels below 80 dBA. These workers were classified based on company records and on duties known to be associated with low noise exposure levels. To control for sex-related differences and potential asymmetry in auditory processing, only male participants whose better ear was the right one (based on auditory thresholds and DPOAEs) were included. All participants were native Mandarin speakers.

2.2 Procedures

A research team member administered a questionnaire to participants in both groups. The questionnaire aimed at gather information on general demographics (e.g., age), occupational history (e.g., workplace type, job description, length of employment, duration of daily noise exposure, and history of hearing protection use), and overall health status (e.g., history of ear disease or ototoxic drug exposure). Participants with a history of ear disease or other health conditions associated with auditory disorders were excluded from the study. Additionally, participants in both groups were asked about significant recreational noise exposure, including firearms use, playing a musical instrument or performing in a band, frequent attendance at concerts or sporting events, visits to noisy bars and or nightclubs, and excessive use of personal listening devices at high volumes. Selected participants were scheduled for an assessment session at Zhejiang Chinese Medical University (Hangzhou, China), where they underwent bilateral otoscopy and the following procedures (see below).

2.2.1 Tympanometry

Tympanograms for both ears were obtained with an Interacoustics Titan device (Middelfart, Denmark) over a pressure range of +200 to −400 daPa with a 226 Hz probe tone. Middle-ear pressure and compliance were recorded. Participants whose tympanograms were classified as anything other than type A in both ears, based on Jerger's classification (1970), were excluded.

2.2.2 Pure-tone audiometry

Air-conduction hearing thresholds were measured using pure tones between 0.25 and 16 kHz with an Interacoustics AC629 (Middelfart, Denmark) clinical audiometer equipped with Sennheiser HDA 300 headphones. The Hughson-Westlake procedure, as described by Carhart and Jerger (1959), was used. Hearing thresholds were obtained at 0.25, 0.5, 1, 2, 3, 4, 6, 8, 9, 10, 11.2, 12, 14, and 16 kHz. Normal hearing was defined as pure-tone hearing thresholds ≤ 20 dB HL for octave and interoctave frequencies between 0.25 and 8 kHz in the tested ear (i.e., right ear), which was set as an inclusion criterion for the study (Jerger and Jerger, 1980). The PTA₄ (average of 0.5, 1, 2, and 4 kHz) and PTA_HF (average of 9–14 kHz) were calculated for analyses purposes.

2.2.3 Measure of occupational noise exposure

For each noise-exposed participant, a single shift-long noise recording was obtained using an ASV5910-R digital recorder (Hangzhou Aihua Instruments Co., Hangzhou, China). The microphone sensitivity level is 2.24 mV/Pa, and the measurement range is 40–141 dBA. Before and after each sampling period, the recorder was calibrated using a sound level calibrator (Hangzhou Aihua Instruments, AWA6221B). During the recording, the device was worn on the worker's shoulder for the entire work shift. The 8-h equivalent continuous sound pressure level (LAeq, 8h) was determined for each worker. Participants with a LAeq, 8h equal to or higher than 80 dBA were classified in the noise-exposed group. In addition, to quantify noise exposure over a year for each noise-exposed worker, a composite noise exposure index, the CNE (in dBA.year), was calculated.

The CNE is defined as: CNE = LAeq, 8h + 10 log T, where LAeq, 8h is the equivalent continuous A-weighted noise exposure level normalized to an 8-h working day (in decibels), and T is the time interval in years.

2.2.4 Distortion product otoacoustic emissions (DPOAEs)

DPOAEs for both ears were obtained, measured, and analyzed using an Interacoustics Titan device with the DPOAE440 module (Middlelfart, Denmark), connected to a Lenovo laptop computer (Beijing, China). DPgrams were obtained with two simultaneously swept primary tones, F1 and F2 (F2/F1 = 1.22), presented at 65 dB SPL and 55 dB SPL, respectively. The primary frequencies selected for the evaluation were the geometric means of F1 and F2 at 1, 1.5, 2, 3, 4, 5, 6, 7, 8, 9, and 10 kHz. The levels of 2F1 – F2 DPOAEs and the corresponding noise floor were recorded as a function of F2, and the signal-to-noise ratio (SNR) was calculated. Participants selected for the study were required to show the presence of DPOAEs (amplitude ≥ −20 dB SPL), and an SNR of at least +6 dB at each of the frequencies in the tested ear (i.e., right ear) from 2 to 10 kHz (Reavis et al., 2011). DPOAE_mean (the average of all frequencies) was calculated for analyses purposes.

2.2.5 Complex-auditory brainstem response (c-ABR)

The c-ABR was recorded using an Intelligent Hearing System (IHS, Smart EP model, Miami, FL, United States) connected to a Lenovo laptop (Beijing, China). Surface electrodes were placed at the midline frontal (Fz, non-inverting electrode), the forehead (Fpz, ground), and at the mastoid (M2, inverting electrode) in accordance with the International 10–20 system of EEG recordings. c-ABR was elicited using a 40-ms synthesized/da/syllable provided by the IHS SmartEP module, delivered via ER3A insert earphones.

The stimulus featured an initial noise burst followed by a voiced formant transition, with a fundamental frequency (F0) increasing linearly from 103 to 125 Hz. Voicing began 5 ms after onset, with a release burst within the first 10 ms. The first formant frequency (F1) increased linearly from 220 to 720 Hz, while the second formant frequency (F2) decreased from 1,700 to 1,240 Hz. The third formant frequency (F3) slightly declined from 2,580 to 2,500 Hz, whereas the fourth (F4) and fifth (F5) formants remained stable at 3,600 Hz and 4,500 Hz, respectively. Despite lacking a steady-state portion, the stimulus is psychophysically perceived as a consonant-vowel syllable. For a detailed description of the synthesized/da/stimulus, see Johnson et al. (2005) and Kraus and Nicol (2005).

The 40-ms/da/stimulus was monaurally presented to the right ear (i.e., better ear) (determined by pure-tone audiometry and DPOAEs) at 80 dB nHL in alternating polarity to minimize stimulus artifact, at a rate of 10.9/s. A time window of 71.81 ms (including a 20 ms prestimulus interval) was used, with an online filter setting of 30–3,000 Hz. Brainstem responses were later bandpass filtered offline from 70 to 2,000 Hz. Trials with artifacts exceeding 30 μV were excluded from the average. A grand average of 6,000 artifact-free responses (two subaverages of 3,000 sweeps) was obtained for each ear. During acquisition, participants were seated in a reclining chair in a quiet, dimly lit room and were instructed to relax with their eyes closed.

For each participant, c-ABR peaks were manually marked following Krizman et al.'s (2019) peak-picking criteria to identify the characteristic seven peaks of the response to the 40-ms/da/stimulus. These included the onset peaks (V and A); the onset of voicing (C), which marks the transition from the aperiodic stop burst to the periodic (voiced) formant transition; the frequency-following response (FFR) peaks (D, E, and F), corresponding to the voiced portion of the syllable; and the offset peak (O) (Johnson et al., 2005; Krizman et al., 2019).

Latencies and amplitudes of individual peaks for the/da/stimulus were further analyzed using an open-source MATLAB-based toolbox developed and distributed by Erika Skoe, Trent Nicol, and Nina Kraus from the Auditory Neuroscience Laboratory, Northwestern University (Skoe and Kraus, 2013, 2010). Using this toolbox, manually selected peak latencies (relative to stimulus onset) and corresponding amplitudes were automatically adjusted (within ±2 sampling points) to obtain the absolute minimum and maximum values.

Further analyses were conducted using the Brainstem Toolbox 2013 across the FFR region of the 40-ms/da/response. The FFR region was defined as the time window between 11.4 and 40.6 ms after stimulus onset, encompassing peaks C, D, E, and F. Spectral encoding across this FFR time window was analyzed using fast Fourier transform (FFT). Average spectral amplitude was calculated for three frequency ranges: the fundamental frequency (F0 amp: 103–120 Hz), the first formant (F1 amp: 455–720 Hz), and a higher-frequency region corresponding to the seventh-to-eleventh harmonics of stimulus F0 (HF amp: 721–1,154 Hz). The root mean square (RMS) amplitude was calculated for the entire duration of the FFR, and the SNR was determined as the ratio of RMS amplitude (signal) to the prestimulus baseline RMS amplitude (noise) (Skoe and Kraus, 2010). A cross-correlation method was applied to determine the stimulus-to-response (SR) correlation. This technique assesses the relationship between neural signals and the stimulus using Pearson's correlation coefficient (r). To identify the optimal temporal alignment, one signal was shifted in time relative to the other to determine the time lag required for maximum correlation.

2.2.6 Measure of speech perception in noise with the Mandarin Bamford-Kowal-bench sentence test (Mandarin BKB)

Speech perception in noise was evaluated using the Mandarin BKB (Xi et al., 2012) in the right ear (i.e., better ear) (as determined by pure-tone audiometry and DPOAEs). A practice round of 10 sentences was first conducted to ensure that participants understood the instructions.

Two lists of 10 sentences were then presented monaurally through HDA 300 headphones (Sennheiser, Germany) at a fixed speech level of 70 dB HL, against a background of four-talker babble noise (three female and one male voices). Each list contained 10 sentences; the first sentence had four key words, while the remaining sentences had three key words each. For each list, SNRs varied from +21 dB to −6 dB, starting with the most favorable SNR (+21 dB) and decreasing in 3 dB steps to more challenging SNRs. Participants were instructed to repeat each sentence aloud. The SNR-50% for one list was calculated by summing the number of correctly repeated key words and subtracting this number from 23.5. This value represents the SNR at which a listener correctly identifies 50% of the key words. The final SNR-50% score was obtained by averaging the SNR-50% values from both lists. A lower SNR-50% score indicates better speech-in-noise perception.

2.3 Statistical analysis

Statistical analyses were performed using Prism V10.3.11 (GraphPad Software, 2024), SPSS V29.0.1.0 (IBM Corp, 2023), and R (R Core Team, 2025). Initially, to compare participants in the noise-exposed and control groups, a two-way ANOVA with repeated measures was performed for audiometric thresholds and DPOAEs. Group comparisons for age, BKB scores, FFR SNR, FFT results, and SR results were then conducted using the Mann-Whitney test. To examine group differences for each latency peak of the c-ABR, a weighted linear regression with propensity score weighting was conducted for each peak, controlling for age. Post-hoc p-value correction for multiple comparisons was performed using the Holm method. To further examine the c-ABR peak latency data, a Bayesian analysis was conducted by regressing group on each latency peak while controlling for age. Weakly informative priors were used (mean = 0, SD = 10). Hamiltonian Monte Carlo sampling was performed via Stan, using 4 chains with 2,000 iterations each (1,000 warm-up). Next, a correlation matrix stratified by group was computed, including age, BKB score, F0 amplitude, PTA₄, PTA_HF, DPOAE_mean, FFR SNR, SR corr, SR Lag, and c-ABR peak C latency. To follow up on the observed correlation and to account for potential confounding by age differences between groups, a weighted linear regression with propensity score weighting was conducted, regressing CNE on BKB while controlling for age. Finally, a multivariable regression model with propensity score weighting for age was constructed to investigate the association between group category (exposed versus non-exposed) and c-ABR peak C latency, while controlling for covariates including age, PTA, and others.

3 Results

3.1 Age, pure-tone thresholds, and DPOAEs

3.1.1 Age

The mean age was 27.14 ± 4.03 years for noise-exposed participants and 23.39 ± 5.13 years for control group participants. Control group participants were significantly younger than noise-exposed participants (p < 0.0001; Figure 1A).

3.1.2 Pure-tone audiometry

As previously mentioned, one of the inclusion criteria was that all participants should have presented with hearing thresholds equal to or better than 20 dB HL in both ears from 0.25 to 8 kHz. For this study, we selected only participants whose better ear was the right one (based on hearing thresholds and DPOAEs). Figure 1D displays the mean hearing thresholds (+/– SEM) in the right ear (0.25–16 kHz) for each group at all tested frequencies. No significant differences (p > 0.05) were found between groups in pure-tone thresholds across the frequency range (0.25–16 kHz) or in the pure-tone average for high frequencies (PTAHF) (p > 0.05; Figures 1C, D). However, a statistically significant difference was observed in the pure-tone average across four frequencies (PTA4) (p = 0.04; Figure 1C).

3.1.3 DPOAEs

Another inclusion criterion was that all participants were required to show presence of DPOAEs (amplitude ≥ −20 dB SPL), and an SNR of at least +6 dB at each of the specified frequencies in the right ear from 2 to 10 kHz. For this study, we selected only participants whose better ear was the right one (based on hearing thresholds and DPOAEs). Figure 1E displays the DPOAEs amplitude and the noise floor in the right ear (1–10 kHz) for each group at all tested frequencies. No significant difference (p > 0.05) in DPOAE amplitudes was found between groups across the frequency range (1–10 kHz).

3.2 Speech in noise

The group mean BKB score for the right ear was 0.14 ± 0.54 for noise-exposed participants and 0.28 ± 0.41 for control group participants. There was no significant difference (p > 0.05) between groups (Figure 1B). This indicates that there was no difference in speech perception in noise between the groups.

3.3 c-ABR

3.3.1 Groups differences for peak latencies

Figure 2 displays the grand average cABR for both groups. Latency values for all the main c-ABR peaks using the 40-ms/da/syllable for the right ear and for both groups are reported in Table 1. To assess group differences for each latency peak, weighted linear regressions with propensity score weighting were conducted for each peak, controlling for age. Post-hoc p-value corrections for multiple comparisons were applied using the Holm method. Results are presented in Table 1. A significant effect of group was found on C-ABR Peak C. Noise-exposed workers showed a significantly longer latency than control group participants (p = 0.01). To further examine the data, a Bayesian analysis was conducted by regressing group on each latency peak while controlling for age (see Table 2). Results showed a distinct group effect for peak C, with a direct negative probability of 1, a 0% ROPE, and a wholly positive credible interval, suggesting a very strong effect for this peak. For the remaining peaks, most values indicate unclear results, except for peaks A and O, which displayed low ROPE values and strong direct probability, even though their confidence intervals include 0.

Table 1. Means, standard deviations (SD), and group comparisons for cABR wave latencies (V, A, C, D, E, F, and O), using weighted linear regression with propensity score weighting controlling for age.

An asterisk (^*) denotes a statistically significant difference between groups based on post-hoc analyses with correction for multiple comparisons.

^*Significant after Holm correction for multiple comparisons.

Table 2. Bayesian regression of group on cABR wave latencies controlling for age: posterior estimates, 95% credible intervals, and probabilities.

B, posterior mean (regression coefficient); SE, standard error; 95% CI, 95% confidence interval; ROPE, region of practical equivalence; P(β < 0): posterior probability that β is less than 0.

3.3.2 Stimulus-to-response (SR) correlations

The SR correlation score and lag are displayed in Figure 3B. The maximum SR correlation did not differ significantly between groups (p > 0.05). The lag between the stimulus and the response, which represents the time shift necessary to obtain the highest correlation, was delayed by 0.27 ms in the control group compared to the noise-exposed group. However, this delay was not statistically significant (p > 0.05).

3.3.3 FFR spectral analyses

Further analyses were conducted on the sustained components of the FFR to examine potential group differences in neural phase-locking to the frequency components. Results are presented in Figure 3A. The SNR magnitude for the 40-ms/da/syllable was not statistically different between groups. Using FFT, we examined spectral amplitude for the frequency region around the fundamental frequency (F0 103–120 Hz), and other formants (F1 455–720 Hz; FHF 721–1,154 Hz). The spectral magnitudes for the fundamental frequency and harmonic components did not show significant differences (p > 0.05) between groups.

3.4 Correlation between noise exposure and auditory outcomes

A Spearman correlation matrix between age, BKB score, F₀ amplitude, PTA₄, PTA_HF, DPOAE_mean, FFR SNR, SR corr, SR lag, and c-ABR peak C latency was obtained for both the control group (Table 3A) and the occupational noise-exposed group (Table 3B), independently. First, a significant correlation was observed between BKB score and CNE in the noise-exposed group. This indicates that the higher the CNE score, the higher the BKB score, suggesting that greater exposure to occupational noise is associated with poorer performance in speech perception in noise (i.e., higher SNR). To follow up on this observed correlation and to account for potential confounding by age differences between groups, a weighted linear regression with propensity score weighting was conducted, regressing CNE on BKB while controlling for age. The results confirmed the relationship (see Table 4 and Figure 4). Several other significant correlations were found in the noise-exposed group: negative correlation between C peak latency and PTA₄, negative correlation between C peak latency and F₀, negative correlation between SR corr and DPOAEmean. In the control group, the following significant correlations were found: negative correlation between C peak latency and SR lag, negative correlation between C peak latency and age, negative correlation between age and BKB score, negative correlation between DPOAE_mean and PTA_HF, positive correlation between PTA₄ and PTA_HF. Finaly, a significant positive correlation was found in both groups between SNR of the neural response and the F₀ magnitude.

Table 3A. Correlation coefficients (Spearman) between BKB, PTA4, PTAHF, Amp F0, DPOAEmean, measures of the FFR, age, C peak latency and cumulative noise exposure (CNE) for the noise-exposed group.

Bold values indicate statistically significant correlations (p < 0.05).

Table 3B. Correlation coefficients (Spearman) between BKB, PTA₄, PTA_HF, Amp F0, DPOAE_mean, measures of the FFR, age and C peak latency for the control group.

^*p < 0.05; ^**p < 0.01; ^***p < 0.001; ^****p < 0.0001.

Table 4. Age-adjusted association between CNE and BKB using propensity score–weighted linear regression.

3.5 Association between group category and c-ABR peak C latency

To investigate the association between group category (noise-exposed and control) and C peak latency, while controlling for covariates such as age, audibility (PTA₄ and PTA_HF), speech perception in noise (BKB score), and OHC integrity (DPOAE_mean), a multivariable linear regression with propensity score weighting for age was conducted to account for group differences. Results are presented in Table 5. The model showed that both group category and age were significantly associated with C-peak latency, while controlling for all other covariates.

Table 5. Multivariable regression analysis for C latency with propensity score–based adjustment for age.

4 Discussion

This study aimed to investigate whether c-ABR outcomes and speech perception-in-noise performance (BKB scores) are associated with occupational noise exposure in young workers with normal hearing thresholds and presence of otoacoustic emissions, a population at high-risk for cochlear synaptopathy.

We examined the onset (V, A) and offset (O) peaks, as well as several sustained components (C, D, E, and F) of the c-ABR elicited by a/da/syllable. While no group differences were observed in the amplitudes of these components, we found a significant delay in the C-peak latency in the noise-exposed group. These results differ from those of Sreedhar et al. (2022) who reported increased latencies for peaks V and A and decreased amplitudes for peaks D and F in a young group exposed to occupational noise.

No group differences in envelope encoding (F0) were observed for the/da/syllable, indicating that occupational noise exposure did not appear to impact participants' ability to phase-lock to the stimulus periodicity. This is consistent with studies by Suresh and Krishnan (2022), Grose et al. (2017), Guest et al. (2018), and Prendergast et al. (2017), which also failed to show a link between noise exposure and envelope encoding. However, our results contrast with those of Bramhall et al. (2021) who found reduced mean EFR magnitude in veterans with significant noise exposure history compared to non-veteran controls. In addition, our findings differ from those in mice with histologically confirmed cochlear synaptopathy. Shaheen et al. (2015) demonstrated a degradation in envelope periodicity and phase-locking value in these mice.

The reasons for the discrepancy between our results and those of the previously cited study remain unclear. It is possible that the speech stimulus used in our study influenced the c-ABR outcomes. Krishnan et al. (2005) demonstrated that speech encoding at the brainstem level is sensitive to language experience. Their study showed that FFR responses to four Mandarin tones were more robust and accurate, with stronger phase-locking activity and higher F2 magnitudes in Mandarin native speakers compared to English speakers. Given that our stimulus did not contain a lexical tone and that our participants were native Mandarin speakers, it is plausible that the stimulus was not sufficiently representative of their language to elicit strong FFR responses.

Another possible explanation relates to our sample selection. In this study, we recruited participants with normal hearing thresholds and normal DPOAE amplitudes, which may have introduced a selection bias, particularly among noise-exposed participants. Several studies have reported differences in individual susceptibility to noise, varying by sex, age, and even among individuals of the same sex and age. It is possible that we inadvertently selected individuals with more resilient hearing, who are less susceptible to the effects of noise exposure and therefore less likely to show hearing deficits related to cochlear synaptopathy (Cody and Robertson, 1983; Lie et al., 2016; Maison and Liberman, 2000). In fact, despite the noise-exposed group having a history of exposure above 80 dBA for at least 2 years, no significant differences in DPOAE amplitudes or hearing thresholds were found between groups, except for the PTA₄, although the difference is not clinically significant (1.05 ± 0.75 dB HL). The inconsistent use of hearing protection in the noise-exposed group suggests that it is unlikely to have played a protective role in preventing auditory deficits. Therefore, the lack of significant differences between groups cannot be easily attributed to a protective effect of hearing protection.

Regarding speech perception in noise, no significant differences in BKB scores were found between groups. This may have been influenced by the choice of speech perception test. Wilson et al. (2007) reported that the Quick Speech-In-Noise (QuickSIN) test and Words-In-Noise (WIN) materials are more sensitive measures of speech perception in background noise than the BKB material. At the time we conducted this study, the only speech-in-noise perception test available to us in Mandarin was the BKB. Thus, the speech-in-noise test used in this study may have not been sensitive enough to find differences between noise-exposed and control group participants. Another possible explanation is that the selected noise-exposed workers may have had more resilient auditory systems, as explained above. As a result, despite their noise exposure levels, their hearing outcomes did not differ significantly from those of the control group, who may not necessarily be less susceptible to the effects of noise.

Nevertheless, even among these potentially more resistant individuals, higher noise exposure levels may have had effect on speech-in-noise perception. This is because within the noise-exposed group, we observed a significant correlation between occupational noise exposure levels and BKB scores, suggesting that greater noise exposure was associated with poorer speech perception in noise (i.e., higher SNR). Considering that all participants had normal audiometric thresholds due to our inclusion criteria, this correlation may suppport the presence of cochlear synaptopathy in the noise-exposed group. The lack of significant differences in BKB scores between the two groups may be explained by unexpectedly poor performance in the control group. Notably, recreational noise exposure was not controlled, and it is possible that some control participants also exhibited signs of hidden hearing loss and cochlear synaptopathy. This could account for the absence of group differences not only in the speech-in-noise test scores but also across all analyzed FFR outcomes. Future longitudinal studies are needed to test this hypothesis and better understand the association between occupational noise exposure and possible signs of cochlear synpatopathy in individuals with normal hearing threholds.

4.1 Limitations

The present study has several limitations. First, we did not employ a detailed questionnaire, such as the NESI or LENS-Q, to assess lifetime noise exposure. Consequently, some noise-exposed workers may have had lower cumulative exposure than certain control participants, which could partly explain the nonsignificant findings. Future studies should incorporate comprehensive lifetime noise exposure assessments to provide a more complete interpretation of results.

Second, CNE was obtained only for noise-exposed participants, preventing group comparisons based on cumulative noise exposure. Moreover, noise dosimetries were measured only once, which may not accurately represent participants' lifetime noise exposure. Future research should consider multiple measurements to better capture long-term exposure.

Third, overall hearing protection was reported to be used inconsistently among noise-exposed workers, and therefore this information could not be included as a covariate in the statistical analysis. Future studies should more accurately capture hearing protection use, as it may influence the interpretation of results.

Fourth, we used the BKB test to assess speech perception in noise. While other tests may be more sensitive, the BKB was the only test available at the time of data collection. Future studies should consider alternative speech-in-noise tests when available.

Finally, the study sample included only Chinese participants, limiting external validity. Previous longitudinal studies have shown that White individuals often have higher prevalence of hearing loss or worse hearing thresholds than other racial groups (Ishii and Talbott, 1998; Dillard et al., 2025), particularly White men (Dillard et al., 2025). Therefore, our findings may not generalize to other racial groups. Future research should include participants from diverse racial backgrounds to improve generalizability and explore potential effects of race on susceptibility to cochlear synaptopathy.

5 Conclusion

In this study, we explored the relationship between occupational noise exposure, c-ABR outcomes, and speech-in-noise perception in young male workers with normal hearing thresholds. Our findings revealed a delayed C-peak latency in the noise-exposed group and a correlation between higher noise exposure levels and poorer speech perception in the presence of background noise. However, the results of this study do not provide strong evidence for a link between occupational noise exposure and cochlear synaptopathy in young male adults with normal hearing thresholds and intact DPOAEs, as assessed using c-ABR and BKB. Further studies are necessary to explore the possible association between noise exposure, c-ABR outcomes and speech perception in noise in young workers exposed to occupational noise.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Zubair Ahmed, University of Maryland, United States

Reviewed by: Adam Kaufman, University of Maryland, United States

Nilesh Washnik, Ohio University, United States