In total, 2,119 workers exposed to silica dust who underwent occupational health examinations between March 2023 and June 2024 were initially recruited. After excluding individuals with other lung, endocrine, or metabolic diseases, and those who have taken antioxidant drugs such as vitamin C 1 month before the physical examination, prior exposure to other toxic agents before employment at the iron mine, and incomplete responses, 1,964 participants met the inclusion criteria, yielding a response rate of 92.69%. Of these, 1,823 did not have silicosis (102 in the silica dust-free control group and 1,721 in the exposed non-silicosis group). The remaining 141 had silicosis (stage I, II, or III), resulting in a prevalence rate of 7.18%. All participants provided written informed consent. The study was approved by the Ethics Committee of North China University of Science and Technology (Approval Number: 2021041).

Sociodemographic data, including sex, age, marital status, and educational level, were collected, along with information on behaviors and lifestyle (e.g., alcohol use and smoking), personal and family medical history, and detailed occupational history (e.g., job type, work location, position, employment date, working hours, and exposure to harmful occupational factors). Physical examination results (e.g., height, weight, lung function, blood tests, and urinalysis) were also recorded. In addition, enterprise-level information such as production processes, environmental monitoring, and general company conditions was collected.

Silicon levels in blood and urine samples were measured using ICP-MS, following procedures detailed in our previous study (6). All operations were conducted under silicon-free conditions.

Total dust concentration, particle dispersion, and free silica content in settled workplace dust were measured using the filter membrane gravimetric method, in accordance with Chinese standard GBZ/T192.1–2007.

Sampling methods of air dust sample adopts a combination of personal sampling and fixed-point sampling. In accordance with the requirements of Chinese “Sampling Specifications for the Detection of Hazardous Substances in the Air of Workplaces,” sampling should be conducted separately based on the type, nature, fluctuation situation and degree of hazard. Dust concentration determination: Set up dust detection points. Each detection point should be measured 1 to 3 times per shift for 2 consecutive days. The workers wore individual dust samplers and measured for two work shifts. Fixed-point sampling was carried out using the AKFC-92A dust sampler (Shanghai Yichang Industrial Co., Ltd. China). The sampling filter membrane was a mixed cellulose ester filter membrane with a diameter of 37 mm and a pore size of 0.8 μm, with a flow rate of 5 L/min. The flow rate of the individual sampler is 2 L/min, collecting air samples from the entire shift of the workers (8 h). As shown in Supplementary Table 1.

Each worker’s dust exposure years were calculated based on the start and end dates (employment time) of each job type and associated exposure to harmful factors recorded in their occupational history. CDE was then estimated by job type using the following formula:

The unit of CDE is mg/m³·years. S_jt is the average annual exposure measurement for job type t in workshop j during year TTH, ds is the year the worker began job type t in workshop j, and de is the year the worker ended certain job type t in workshop j (For silicosis patients, de is the year diagnosed with silicosis). CDE reflects both the duration and intensity of dust exposure (10).

Statistical analyses were conducted using SPSS 23.0 and SAS 9.4. Normally distributed quantitative data are expressed as mean ± standard deviation, while categorical variables are described as composition ratios. Pearson correlation, Spearman correlation, and chi-square test were used for correlation analyses of normally distributed, non-normally distributed, and categorical data, respectively. ROC curves and RCS were used to identify screening thresholds. All hypothesis tests were two-sided, with a significance level of α = 0.05.

All participants were male and were classified into non-silicosis and silicosis groups based on disease status. Potential risk factors for silicosis in dust-exposed workers, including age, smoking, alcohol consumption, body mass index (BMI), and forced expiratory volume in 1 s (FEV1%), were described and compared. Significant differences were observed between the two groups regarding age, drinking, BMI, FEV1%, and years of dust exposure (p < 0.05), while no difference was found in smoking status (Table 1).

The primary job categories in the mine include mining, crushing, grinding and selection, as well as auxiliary operations. Based on dust exposure data from 2021 to 2024, an analysis of variance revealed significant differences among job types, except for auxiliary roles (p < 0.01; Table 2).

CDE, blood silicon, and urine silicon levels were compared among the dust-free control, dust exposure, and silicosis groups. Analysis of variance revealed significant differences across all groups (p < 0.01; Table 3).

Correlation analysis was conducted between internal exposure indicators (blood and urine silicon levels) and the external exposure indicator (CDE). Both blood and urine silicon levels showed a significant positive correlation with CDE (p < 0.01). The correlation with urine silicon was strong, while that with blood silicon was moderate (Table 4).

Blood and urine silicon levels (internal exposure), CDE (external exposure), and years of dust-exposed work were selected as early screening markers for silicosis. Thresholds for each screening index were identified using RCS and ROC curves. According to the relationship curves, blood and urine silicon levels rose rapidly within the first 1–6 years of exposure, stabilized between years 6 and 10, and then declined. Urine silicon levels peaked around year 9, while blood silicon levels peaked around year 10. These findings suggest that the highest risk period for developing silicosis occurs between 6 and 10 years after initial silica-dust exposure (Figure 1).

RCS and the ROC curves were jointly used to determine thresholds for blood and urine silicon as screening indicators for silicosis. Since RCS did not identify clear thresholds, only the points corresponding to the maximum kappa values from the ROC curves were selected.

For blood silicon, the ROC analysis showed statistical significance (p < 0.05), with an area under the curve (AUC) of 0.557. Analysis showed that the optimal blood silicon threshold was 8.015 μg/L, corresponding to the highest kappa value, with sensitivity and specificity of 0.776 and 0.455, respectively (Table 5; Figure 2A).

For urinary silicon, ROC analysis revealed statistical significance (p < 0.05), with an AUC of 0.647. The optimal urinary silicon threshold was 9.51 μg/L, yielding the highest kappa value, with sensitivity and specificity of 0.917 and 0.444, respectively (Table 6; Figure 2B).

The AUC of both blood silicon and urine silicon is relatively low, especially blood silicon with 0.56 AUC is marginal at best. Although it within the scope of statistics, external validation is still required to obtain the corresponding threshold.

The ROC curve and RCS were used to determine the threshold for CDE as a screening indicator for silicosis, yielding values of 3207.235 mg·years and 4,250 mg·years, respectively. The average threshold was 3728.5 mg·years. Accordingly, the CDE among workers in this iron mine should not exceed 3728.5 mg·years.

The ROC curve was used to determine the threshold of CDE as a screening indicator for silicosis. The results were statistically significant (p < 0.05) with an AUC of 0.658, indicating acceptable screening performance. The optimal CDE threshold, corresponding to the kappa value, was 3207.235 mg·years, with sensitivity and specificity of 0.569 and 0.735, respectively (Table 7; Figure 3).

Using an RCS model with four knots, the Akaike Information Criterion (AIC) of the model was 1090.70. CDE showed a nonlinear dose–response relationship with silicosis. The overall association was significant (χ² = 102.54, p < 0.01), as was the test for nonlinearity (χ² = 71.62, p < 0.01). Based on the odds ratio (OR) confidence interval, CDE between 1750 and 7,250 mg·years was associated with an increased risk of silicosis (OR >1). The risk peaked at approximately 4,250 mg·years, indicating a rapid increase in risk from 1750 to 4,250 mg·years, followed by a gradual decline between 4,250 and 7,250 mg·years (Figure 3). This phenomenon is quite interesting and seems contrary to common sense, it should be that the OR value increases with the increase of the CDE. However, as shown in this chart, the OR value does indeed decline after exceeding the peak of 4,250 mg·years. It is speculated that this is related to the susceptibility to silicosis. Before the peak, the risk of developing the disease is relatively high, while after the peak, since those who should have developed the disease have already done so, the remaining population is less susceptible to silicosis. Therefore, the OR value keeps decreasing. This phenomenon also deserves further study in the future.

Using CDE as a screening indicator for silicosis, thresholds identified by the ROC curve and RCS were 3207.235 and 4,250 mg·years, respectively. The average value of 3728.5 mg·years was adopted as 3728.5 mg·years as the screening threshold.

Years of dust exposure are a simple, practical, and effective indicator for early screening of silicosis. Therefore, they were evaluated as a screening threshold using both the ROC curve and RCS, which yielded thresholds of 8.915 and 8 years, respectively. The average, 8.5 years, was adopted as the screening threshold. Accordingly, dust exposure among workers in this iron mine should not exceed 8.5 years.

The ROC curve was used to identify the threshold of dust exposure years for silicosis screening. Results showed statistical significance (p < 0.05) with an AUC of 0.689, indicating acceptable screening performance. The optimal dust exposure threshold was 8.915 years, corresponding to the kappa’s highest value, with sensitivity and specificity of 0.73 and 0.54, respectively (Table 8; Figure 4).

Using the RCS model with four nodes, the AIC of the model was 1094.47. Dust exposure years showed a nonlinear dose–response relationship with silicosis. The overall association was significant (χ² = 82.50, p < 0.01), as was the test for nonlinearity (χ² = 10.19, p < 0.01). When the dust exposure period exceeds 8 years, the OR is greater than 1, indicating increased risk of silicosis. The OR remained stable between 8 and 15 years, then rose sharply beyond 15 years, suggesting a higher risk of silicosis (Figure 4).

In total, 330 individuals were included in the validation set: 79 from a coal mine control group, 112 silica-exposed tunneling workers, and 139 cases of silicosis. ROC curve analysis of blood and urine silicon levels was conducted to assess the accuracy of previously identified screening indicators. Results showed that the blood silicon threshold slightly differed from that of iron mine workers but remained a relatively reliable screening marker for silicosis. In contrast, urine silicon levels showed greater variability, and their threshold requires further validation (Table 9).

Validation of the blood silicon threshold using the ROC curve showed statistical significance (p < 0.05), with an AUC of 0.63. The optimal threshold, based on the highest kappa value, was 8.81 μg/L, with sensitivity and specificity of 0.99 and 0.37, respectively. This closely aligns with the previously identified threshold of 8.02 μg/L for iron ore dust-exposed workers, supporting the reliability of blood silicon as a screening marker for silicosis (Figure 5A).

Validation of the urine silicon threshold using the ROC curve showed significance (p < 0.05), with an AUC of 0.70. The optimal threshold, based on the highest kappa value, was 11.05 μg/L, with a sensitivity of 0.63 and specificity of 0.97. This value is notably higher than the 9.51 μg/mL threshold identified in iron ore dust-exposed workers, indicating that further validation is needed for urine silicon as a screening marker (Figure 5B).