This study adopted a cross-sectional design. Participants were recruited in 2023 from two automotive manufacturing factories in Guangzhou, China. Occupational noise is the main occupational hazard in both factories. Workplaces exposed to occupational hazards during production include welding and stamping workshops, among others, with detailed information can be found in a previous study(Yu et al. 2025). The inclusion criteria were: (1) employment duration of ≥ 1 year; (2) male workers; and (3) provision of informed consent. The exclusion criteria were: (1) missing data on exposure or outcome and (2) a history of physician-diagnosed liver disease, including chronic hepatitis, hepatitis B, and hepatitis C, cirrhosis, or liver cancer. Given the underrepresentation of female workshop workers (approximately 1%), most of whom held office positions, female participants were excluded from this study.

This investigation was carried out in strict accordance with the Declaration of Helsinki and was approved by the ethical review board of the Guangzhou Twelfth People’s Hospital with the certificate number of 2,023,055. All participants provided informed consent prior to their involvement in the original research.

Blood samples were collected from all participants after a minimum of 8 h of overnight fasting. In this study, serum ALT, AST, and GGT were selected as biomarkers of liver function. Serum concentrations of ALT, AST, and GGT were determined using the standardized method recommended by the International Federation of Clinical Chemistry (IFCC) (Schumann et al. 2002a, b, c). Elevated ALT was defined as ALT > 40 U/L (Siddiqui et al. 2013). Elevated AST was defined as AST > 40 U/L (Xie et al. 2019), and elevated GGT was defined as GGT > 50 U/L (Miyake et al. 2003).

Occupational noise exposure was assessed using self-reported questionnaires and occupational monitoring reports. Noise exposure levels were evaluated based on the equivalent continuous A-weighted sound pressure levels (Lex, 8 h) in decibels [dB(A)], following the guidelines of the Classification of Occupational Hazards in the Workplaces [GBZ/T 229.4–2012] (China 2012). Measurements were conducted using a sound pressure-type meter (KSL TECHNOLOGY, KSL-dB2) in accordance with the Measurement of physical factors in the workplace (GBZ/T 189.8–2007, China, 2007). The measurement locations were established based on relevant regulations, and the noise exposure posts were classified accordingly. Each measurement lasted for two hours, and the mean noise level was subsequently calculated. Personal noise exposure was assessed using individual monitoring devices for roles with irregular noise intensity, while fixed-point measurements were employed for positions with stable noise levels. Time-weighted average noise levels were calculated for each job role. Workers were linked to fixed measurement points based on staff records provided by the occupational health department, ensuring individual-level noise exposure data for all participants. The specific process of occupational noise exposure measurement followed the methodology outlined in our previously published studies (Yu et al. 2025). Cumulative noise exposure (CNE) was calculated as the product of the occupational noise exposure level (LAEq,8 h in dB(A) and the duration of exposure (T, in years), using the formula (Stokholm et al. 2013; Wang et al. 2022):

$$} = }0log\left[ }0^}\left( } \right)/}0}} \times T} \right)} } \right] $$

Age, body mass index (BMI), smoking status, alcohol consumption, education, physical activity, income (Yuan/month), hearing protection devices use, shift work, night work, hypertension, diabetes, total cholesterol (TC), triglyceride (TG), high-density lipoprotein (HDL) were obtained by questionnaire and physical examinations. MAFLD was defined as hepatic steatosis with at least one of the following criteria: overweight or obesity, type 2 diabetes, or metabolic risk abnormalities (Eslam et al. 2020; Gofton et al. 2023). BMI was calculated as weight (kg) divided by height squared (m²) based on physical examination data. Current smokers were defined as individuals who smoked ≥ 1 cigarette/day during the past six months. Current alcohol consumers were defined as those who consumed alcohol ≥ 1 time/week during the past year. Education level was categorized as high school or below and college or above. Physical activity was defined as engaging in ≥ 150 min of moderate-intensity or ≥ 75 min of vigorous-intensity per week. Income (Yuan/month) was grouped into three categories: ≤3000, 3001–8000, and ≥ 8000. Hearing protection device use was defined as regular or nearly daily use. Shift work and night work were binary variables, with ‘yes’ indicating participation in a two-shift or three-shift rotation and working hours between 0:00 and 5:00 am, respectively. Hypertension was defined based on (1) self-reported use of anti­hypertensive medications or physician-diagnosed hypertension or (2) measured systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg. Diabetes was defined as fasting plasma glucose (FPG) ≥ 7.0 mmol/L, self-reported use of hypoglycemic medications, or physician-diagnosed diabetes. Serum levels of TC, TG, and HDL were measured by trained professionals using standardized methods.

A descriptive analysis was conducted for the covariates, noise exposures, and outcomes across the entire sample. The distributions of continuous variables were assessed using the Q-Q plots and the Shapiro-Wilk normality test. Normally distributed variables were expressed as mean with standard deviation, while non-normally distributed continuous variables were expressed as medians with interquartile range (IQR). Categorical variables were presented as numbers (percentages). Analysis of variance (ANOVA) was conducted for normally distributed continuous data, while the non-parametric Wilcoxon rank-sum test was used for non-normally distributed continuous data. The Chi-square tests were applied to categorical data to compare baseline characteristics among different liver enzyme groups.

Multivariate linear models were employed to estimate the associations between occupational noise exposure and ALT, AST, and GGT, with regression coefficients (β) and 95% confidence intervals (CIs) calculated. Logistic regression models were used to examine the association between occupational noise exposure and elevated ALT, AST, and GGT, with odds ratio (OR) and 95% CIs determined. Based on previous research, a Directed Acyclic Graph (DAG) was constructed for covariate selection, as shown in Fig. S1.

Two models were fitted: (1) an unadjusted model and (2) a covariate-adjusted model. To explore potential effect modification, subgroup analyses were conducted based on binary categorization of factors, including smoking (yes vs. no), alcohol consumption (yes vs. no), physical activity (yes vs. no), hearing protection device (yes vs. no), shift work (yes vs. no), night work (yes vs. no). Interaction terms were tested at a significance level of P < 0.05. The correlation between CNE, liver enzymes, MAFLD, and WBC was assessed using Pearson’s coefficient and point-biserial correlation coefficient. Furthermore, to evaluate the mediating effects of MAFLD or WBC in the relationship between CNE and liver enzymes, total effect (TE), direct effect (DE), and indirect effect (IE) were estimated, adjusting for covariates. Additionally, multicollinearity was assessed using the variance inflation factor (VIF) and tolerance values.

All statistical analyses were conducted using R software, version 4.3.1. All statistical tests were two-tailed, with statistical significance set at P < 0.05.