2.1 Study subjects and design
A cross-sectional epidemiology study was conducted at Sinopec Maoming Petrochemical Plant (Maoming, China) from October to December 2020. A total of 1496 workers were recruited based on the following criteria: 1) Noise [LEX, 8h ≥ 80dB(A)] exposure for more than one year; 2) TEXS exposure but to no other chemicals or pollutants that could impair hearing function, such as n-hexane; 3) no pathological changes in the external ear canal by otoscopy; 4) no self-report and/or diagnosed auditory system diseases; 5) no use of ototoxic medication, such as gentamicin, streptomycin, kanamycin, i.e., during the past two weeks; 6) completed pure tone audiometry and designed questionnaire including sex, age, lifestyle habits, smoking, and drinking information. The study protocol (No.2020118) was approved by The Ethics Committee of Medical Research, School of Public Health, Sun Yat-sen University. All the participants signed an informed consent in the questionnaire interview.
2.2 Noise exposure assessment and cumulative noise exposure
According to the National Occupational Health Standards of China: Measurement of noise in the workplace (GBZ/T 189.8–2007), a personal noise dosimeter (Casella dBadge2®, UK) was used to measure the noise level during the working day time in selected workplaces for three consecutive days once a year. The normalization of equivalent continuous a-weighted sound pressure level to a nominal 8 hours working day (LEX, 8h) was used to estimate the cumulative noise exposure [CNE, dB(A)·year] for each participant based on the monitoring data of the petrochemical plant in 2019 and 2020. CNE is a comprehensive index combining noise intensity with exposure duration and calculated by the formula in Acoustics — Estimation of noise-induced hearing loss (ISO 1999:2013):
$$\text{C}\text{N}\text{E}={L}_{EX,8h}+10\text{log}T$$
where CNE is cumulative noise exposure and used to quantify the noise exposure for each subject in dB(A)·year; \({L}_{EX,8h}\) is the normalization of equivalent continuous a-weighted sound pressure level to a nominal 8 hours working day in dB(A); and T refers to the duration of exposure to noise in years.
2.3 BTEXS exposure assessment
The concentration of BTEXS in the workplace in 2019 and 2020 was used to group. Air samples were collected by personal/area sampling during the work period (8 hours) according to the national occupational health standards of China: Specifications of air sampling for hazardous substances monitoring in the workplace (GBZ 159–2004). All air samples were measured for BTEXS under national occupational health standards of China: Determination of toxic substances in workplace air - Part 59: volatile organic compounds (GBZ/T 300.59–2017), Part 66: Benzene, toluene, xylene and ethylbenzene (GBZ/T 300.66–2017), and Part 68: Styrene, methyl styrene and divinyl benzene (GBZ/T 300.68–2017). In a word, BTEXS was sampled by drawing workplace air through the active carbon tubes. All samples were sent to the laboratory immediately and analyzed by gas chromatography with a flame ionization detector (GC-FID, Agilent, America) following elution with sulfur dioxide. At least 3 blank samples and 3 repeated samples for each sampling were measured to ensure the quality of the operation processes. The coefficient of determination (R2) of the calibration curve with eight levels of concentrations was more than 0.99 for each BTEXS. The limits of detection (LOD) of BTEXS were 0.030, 0.030, 0.030, 0.030, and 0.040 mg/m3, respectively. Then, the 8-h time-weighted average concentration was calculated based on the worker’s exposure time. All BTEXS below the LODs were defined as "un-exposure", and levels of BTEXS above the corresponding LODs were defined as "exposure".
2.4 Pure-tone audiometry and hearing assessment
We used pure-tone audiometry to evaluate the hearing thresholds according to the national standards of China: Acoustics-Audiometric test methods-Part1: Pure-tone air and bone conduction audiometry (GB/T 16296.1–2018/ ISO 8253-1:2010). The audiometry was performed by physicians from Guangdong Province Hospital for Occupational Disease Prevention and Treatment using a clinical audiometer (Piano Plus, INVENTIS, Italy). All the testing equipment was professionally calibrated. The hearing thresholds for air conduction were determined at 500, 1000, 2000, 3000, 4000, and 6000 Hz. The test was preceded in an audiometric booth where the noise level was below 25 dB(A) and at least 48 hours after the last exposure to noise. The results of audiometry were displayed as pure-tone average (PTA), which is a threshold value measured at a specified set of frequencies. In this study, hearing loss (HL) was defined as the PTA of more than 20 decibels (dB) at 500, 1000, 2000, and 4000 Hz in the better hearing ear (WHO 2021). Meanwhile, we defined speech-frequency hearing loss (SFHL) as the PTA of more than 20dB at 500, 1000, and 2000 Hz in the better ear. High-frequency hearing (HFHL) was also defined as the PTA of more than 20dB in the better ear at 3000, 4000, and 6000 Hz.
2.5 Covariates
We selected covariates a priori through a literature search(Chang et al. 2006; A. Fuente et al. 2013; Kuang et al. 2019; Wang et al. 2019). For HL, SFHL and HFHL, we included smoking status (smoking, non-smoking), drinking status (drinking, non-drinking), BMI (continuous), type of headphone/earphone (headphone, earpieces, earphone, and unknown type), the volume of the headphone/earphone (< 50%, 50 ~ 80%, > 80%), duration of using headphone/earphones (continuous), wearing ear protection (Yes or No), and using personal protective equipment (facial masks/working clothes/gloves, Yes or No) in all analysis models. However, we excluded age and sex in these models because the hearing loss assessment was adjusted for these factors.
2.6 Statistical analysis
We described the distribution of demographics, BTEXS levels, noise levels, and other covariates among the overall study participants. \(\left(\stackrel{-}{x}\pm sd\right)\)or M (P25, P75) was used for continuous variables and counts (percentages) were used for categorical variables.
(1) Logistics models
Toluene, styrene, xylene, and ethylbenzene (TEXS) were usually recognized as ototoxic solvents (Gagnaire and Langlais 2005; Nies 2012), but benzene was not (Gagnaire and Langlais 2005). Therefore, logistic regression models were used to examine the associations between CNE and TXES exposure and hearing loss. We first used single-pollutant models to assess the relationship between CNE and TEXS and hearing loss. For the CNE analysis, participants were divided into four groups following the quartile of CNE levels, with the lowest quartile serving as the referent group. We classified the participants exposure to one individual or more TEXS as the TEXS exposure group and those un-exposure to none of the individual TEXS as the no-TEXS exposure group. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. P-trend values were calculated by incorporating the median of each CNE quartile in the model as a continuous variable.
To exclude the impact of potential confounders, we used three models. Model 1 was an unadjusted crude model. Model 2 was adjusted for smoking status, drinking status, BMI, and using personal protective equipment (hearing protection, facial mask, working clothes, gloves). Model 3 was further adjusted for using headphone/earphone (types, volume, and duration). Then, we further verified the correlation between noise/TEXS and hearing loss using two-pollutant models including both CNE and TEXS, with the same covariates.
We constructed additional logistic regression analyses to explore the joint effects of noise and TEXS on hearing loss by including both noise levels and TEXS levels. That is, we divided the participants into the high-CNE group and low-CNE group on the basis of the median of CNE. Then, all participants were re-classified into four groups, such as “low-CNE + no-TEXS”, “low-CNE + TEXS”, “high-CNE + no-TEXS” and “high-CNE + TEXS”. Odds ratios were estimated to observe joint effects. Then, multiplicative joint effects were also assessed by adding interaction terms of CNE (quartered) and individual TEXS exposure (dichotomous) to the final logistic models. In addition, the models were additionally adjusted by three other homologs to observe the interaction effects between individual TEXS and noise, respectively. Moreover, participants were divided into different subgroups on the basis of each TEXS exposure, and odds ratios were estimated. Due to the small number of participants suffering SFHL, subgroup analyses for SFHL were not performed.
(2) Sensitive analysis
We performed the sensitivity analysis in our study. Firstly, we introduced benzene into the logistics regression model because benzene frequently co-exists with TEXS in the air of petrochemical plant workplaces. The participants exposure to benzene in the no-TEXS exposure group were re-classified to the BTEXS exposure group, and the rest participants in the no-TEXS exposure group were in the no-BTEXS exposure group. We assessed the association between BTEXS and hearing loss. And more, we repeated the logistics regression models with three different models. Model 1 didn’t adjust for any covariates (crude); model 2 was adjusted for smoking status, drinking status, BMI, and using personal protective equipment; and model 3 was further adjusted for using headphone/earphone besides the covariates in model 2.
All statistical analyses were performed by Statistical Product and Service Solutions (SPSS, version 25.0). A two-sided statistical significance level was set at α = 0.05. The artwork was created using GraphPad Prisma 8.3.0.