A cross-sectional occupational-exposure study was conducted in traditional blacksmith workshops in Bandung Regency, West Java Province, Indonesia; a detailed description of the sampling location is available in our previous publication^[22]. The workshops operate in semi-enclosed, naturally ventilated structures. The standard working day was 6 h for blacksmiths (5 h for Forgers) and 8 h for controls, with exposure frequencies of 311 days/year for blacksmiths and 365 days/year for controls, reflecting occupational and residential exposure scenarios, respectively [Supplementary Table 1]. Ethical approval was obtained from the Research Ethics Committee of Universitas Padjadjaran, Bandung, Indonesia (approval number 1151/UN6.KEP/EC/2024, issued 01 November 2024). All ethical requirements were strictly observed to protect participants’ rights and welfare, and informed consent was obtained through a two-stage process combining verbal explanation and signed handwritten consent prior to data collection.

Thirty male blacksmith workers were enrolled across five job categories: Welder (n = 7), Grinder (n = 9), Finisher (n = 4), Forger (n = 7), and Cutter (n = 3). A non-metalworking control group of ten individuals was recruited concurrently from community occupations, including Farmer (n = 5), Tailor (n = 2), Seller (n = 1), Teacher (n = 1), and Driver (n = 1). Controls were selected using inclusion and exclusion criteria comparable to those applied to the exposed group. Eligible controls were individuals who were not blacksmiths, had never engaged in routine metalworking throughout their lives, and resided at least 5 km from any blacksmith workshop or other relevant emission source. No respiratory protective equipment (masks, respirators, or face shields) was worn by any participant during sampling; consequently, the measured breathing-zone concentrations represent the actual inhaled concentrations, without adjustment for a protection factor.

Personal airborne PM was collected using a Personal Nano Sampler (PNS), originally developed by Furuuchi et al. (2010)^[33] and subsequently refined by Thongyen et al. (2015)^[34] for breathing-zone nanoparticle exposure assessment. Sampling was carried out between 17 May and 19 June 2025 in West Java, Indonesia, during the transition to the early dry season. Each worker wore the PNS with the inlet positioned in the breathing zone, approximately 15-20 cm from the mouth and nose, to capture representative occupational inhalation exposure over a full working shift. PM was collected on 46.2 mm PTFE membrane filters (2 µm pore size; Whatman™, Cytiva, UK; Product No. 7592-104) supported by a polypropylene ring; filters were pre-weighed and acid-cleaned before sampling.

A portable battery-operated pump (ASP-6000; Komyo Rikagaku Kogyo K.K., Kawasaki, Japan) was connected to the PNS outlet via flexible resin tubing and operated at a flow rate of 4 L/min, with flow verified before and after each sampling event. Samples collected on the impactor and inertial-filter substrates, corresponding to the two aerodynamic size fractions PM_0.5 and PM_0.5-1, were selected for gravimetric and chemical analysis because these ranges encompass the respirable fraction most relevant to deep-lung deposition.

Filters were conditioned at 20 ± 2 °C and 45% ± 5% relative humidity (RH) for at least 48 h both before and after sampling to equilibrate moisture content. Pre- and post-sampling weights were determined on a Mettler Toledo MX5 semi-microbalance (Mettler Toledo, Switzerland; readability 1 µg) under these controlled conditions. Mass concentrations of PM in each size fraction were calculated from the net mass gain and the total volume of sampled air. Given the exceptionally high PM loadings recorded in Grinder samples, all post-sampling filters were visually inspected; they appeared homogeneous, with no evidence of overloading, particle breakthrough, or membrane rupture, and all samples were therefore retained for analysis.

The chemical composition of the filter-collected PM was determined on an Epsilon 5 energy-dispersive X-ray fluorescence (EDXRF) spectrometer (PANalytical, The Netherlands) at the National Research and Innovation Agency (BRIN), Bandung, Indonesia, following US EPA Compendium Method IO-3.3 (Determination of Metals in Ambient Particulate Matter Using X-Ray Fluorescence Spectroscopy)^[35]. The instrument is equipped with an Sc/W X-ray tube and a high-resolution Ge detector, and was calibrated with thin-film standards (Micromatter Technologies, P220-Prague and V149-Vienna) using least-squares regression. Quality assurance and method validation were performed using NIST SRM 2783 (Air Particulate on Filter Media). Recoveries for all sixteen analyzed elements ranged from 98%-102%, with a systematic bias of -1.99% to +1.99% and analytical precision (%CV) of 0.60%-7.23%, indicating good accuracy and reproducibility. A blank membrane filter was analyzed and used for blank subtraction across all samples, and QC samples were monitored periodically throughout each analytical run to verify instrument performance. All sixteen target elements were consistently detected above background levels in every worker sample.

XRF is a nuclear-based analytical technique that provides both qualitative and quantitative elemental determination by measuring the characteristic X-rays emitted from atoms excited by a primary X-ray source. It was selected here because it is non-destructive, preserving the samples for further analysis, and because it enables rapid multi-element detection with minimal sample preparation; the method is widely used in PM characterization studies. Sixteen elements were quantified: aluminum (Al), silicon (Si), sulfur (S), chlorine (Cl), potassium (K), calcium (Ca), titanium (Ti), chromium (Cr), manganese (Mn), iron (Fe), cobalt (Co), nickel (Ni), copper (Cu), zinc (Zn), arsenic (As), and lead (Pb). The airborne concentration of each element (C_A, µg/m³) was calculated by dividing the measured elemental mass on the filter by the total volume of sampled air. The EDXRF scan covered the full active filter area, so concentrations were reported directly without area extrapolation. Because the method determines total elemental concentrations and cannot distinguish chemical species or oxidation states [e.g., Cr(III) vs. Cr(VI), or Mn²⁺ vs. Mn³⁺/Mn⁴⁺], all health-risk characterization is based on total elemental concentrations. This is a conservative assumption and is consistent with standard practice in occupational health-risk assessment when speciation data are unavailable.

CA denotes the elemental concentration measured directly in the breathing zone over full working shifts. Because no participant wore respiratory protective equipment during sampling (see previous section), CA represents the actual inhaled concentration without adjustment for a protection factor. For contextual comparison with occupational exposure limits expressed as 8-hour time-weighted averages (TWA), CA values were scaled as CA × (ET/8) for groups with ET < 8 h. This scaling is provided for reference only and does not affect the primary risk calculations.

Individual EC were calculated following the Risk Assessment Guidance for Superfund, Volume I: Human Health Evaluation Manual (Part F, Supplemental Guidance for Inhalation Risk Assessment; US EPA, 2009)^[36]:

(1) $$ \mathrm{EC=(CA\times ET\times EF\times ED)/AT} $$

where EC = exposure concentration (µg/m³), CA = contaminant concentration in air (µg/m³), ET = exposure time (h/day), EF = exposure frequency (days/year), ED = exposure duration (years), and AT = averaging time (h). For the carcinogenic pathway, AT_car = 70 yr × 365 days × 24 h = 613,200 h for all participants; for the non-carcinogenic pathway, AT_noncar = ED × 365 × 24 and therefore varies by individual. Two EC values were computed per participant per metal, namely EC_NonCar (non-carcinogenic) and EC_Car (carcinogenic).

Non-carcinogenic risk was calculated as HQ = EC_NonCar/RfC, and the HI as HI = ΣHQ. An HI ≤ 1 indicates acceptable risk, whereas HI > 1 indicates potential adverse non-carcinogenic effects. Inhalation reference concentrations (RfC, µg/m³) compiled from US EPA (2023), US EPA (2024), ATSDR (2021), NAAQS (2022), OSHA, and OEHHA (2024) were applied for the eight metals with adequate data: Cr (1,000), Mn (0.05), Co (0.02), Ni (0.014), Cu (100), Zn (5,000), As (0.015), and Pb (0.15) [Supplementary Table 2]^[36].

Carcinogenic risk was calculated as ILCR = EC_Car × IUR, with ILCR_total = ΣILCR. An ILCR between 1 × 10^-6 and 1 × 10^-4 is considered acceptable, whereas ILCR > 1 × 10^-4 exceeds the upper threshold. Inhalation unit-risk [IUR, (µg/m³)^-1] values from the US EPA IRIS database were applied for Co (9 × 10^-3), Ni (2.4 × 10^-4), As (4.3 × 10^-3), and Pb (8 × 10^-5) [US EPA IRIS, 2024; and US EPA Regional Screening Levels (RSL), 2024]^[36].

Descriptive statistics [mean, standard deviation (SD), median, P95, and maximum] were computed by job category. The Shapiro–Wilk test confirmed that all key variables were non-normally distributed. Kruskal–Wallis H tests assessed inter-group differences across the six groups (five worker categories plus controls), and two-sided Mann–Whitney U tests compared pooled workers with controls. Statistical significance was set at P < 0.05.

PM concentration data are presented in Supplementary Table 3 and Figure 1, and participant characteristics with job-specific exposure parameters are summarized in Supplementary Table 1. Mean PM_0.5 concentrations ranged from 193 µg/m³ (Cutters; SD 121.3) to 963 µg/m³ (Grinders; SD 1,152). For PM_0.5-1, Grinders dominated, with a mean of 6,945 µg/m³ (SD 14,201) and a maximum of 44,116 µg/m³, exceeding all other groups by a factor of four or more and reflecting the intense particle generation associated with mechanical abrasion. Total PM (the sum of PM_0.5 and PM_0.5-1, i.e., PM₁) was highest in Grinders (7,908 ± 15,258 µg/m³) and Welders (2,507 ± 1,590 µg/m³), whereas controls averaged only 101 µg/m³ (SD 35.7). The Kruskal–Wallis test confirmed significant inter-group differences (H = 30.8, P < 0.001). Individual job-category means exceeded the control mean by 25- to 78-fold (Welders and Grinders, respectively), while pooled workers as a group exceeded controls by a median factor of ~14 (median 1,467 vs. 103 µg/m³; Mann–Whitney U = 300, P = 0.000003).

The PM concentrations documented here are extreme by any international standard. Because no dedicated standard exists for any PM size fraction in indoor or occupational settings, ambient air-quality standards are commonly used as the reference; among these, PM_2.5 guidelines are most frequently applied because their health effects are well established and widely recognized by bodies such as the World Health Organization (WHO). Since this study measured PM₁ (the sum of PM_0.5 and PM_0.5-1), PM₁ concentrations were compared with PM_2.5 standards as the closest available reference. In our previous study^[22], PM1 accounted for 85%-92% of PM_2.5, indicating that the great majority of particles are sub-micron and only about 8%-15% exceed 1 µm^[22]. For example, the mean PM₁ concentration for Grinders (7,908 µg/m³) exceeds the Indonesian National Ambient Air Quality Standard (NAAQS; 55 µg/m³, 24-h PM_2.5) by a factor of 144, the US EPA standard (35 µg/m³, 24-h PM_2.5) by a factor of 226, and the WHO 24-h guideline (15 µg/m³ PM_2.5) by a factor of 527^[37,38]. Welder total PM (2,507 µg/m³) exceeds the NAAQS 46-fold^[39].

The markedly higher PM_0.5-1 concentration in Grinders (mean 6,945 µg/m³) relative to their PM_0.5 concentration (963 µg/m³) indicates that grinding predominantly generates coarser particles. This is consistent with the mechanism of grinding, in which mechanical fracture of the abrasive wheel and metal surface produces particles typically in the 0.5-10 µm range. Welding, by contrast, generates much finer particles (< 0.5 µm) through the condensation and nucleation of metal fumes^[40-44]. These contrasting sub-fraction profiles carry direct health implications: welding-fume particles deposit predominantly in the alveolar region and carry a high toxic-metal load per unit mass, whereas grinding particles show mixed tracheobronchial–alveolar deposition and produce total mass exposures far exceeding those of welding.

Cross-country comparisons must be interpreted cautiously because of differences in sampling methods and the particle-size fractions measured. Even so, the available evidence indicates that the PM concentration observed here, particularly for grinders and welders, is substantially higher than those reported elsewhere. For instance, PM_2.5 concentrations in welding environments have been reported at roughly 132-306 µg/m³ in China^[7] and up to 490-1,727 µg/m³ in metalworking settings in Kazakhstan^[45], while in the United States PM₁ and PM_2.5 ranged from 109-196 and 112-197 µg/m³ in small fabrication shops^[46]. By comparison, the mean PM₁ concentration for Grinders in this study reached 7,908 µg/m³ and total PM for Welders reached 2,507 µg/m³, clearly exceeding all previously reported ranges.

Concentrations of all sixteen elements (CA) are reported in Supplementary Table 4 and visualized in Figure 2 and Supplementary Figure 1. Iron (Fe) was the dominant metal in every blacksmith group, followed by Si and K. Welders recorded the highest CA_Fe (380 µg/m³; SD 334), followed by Grinders (200 µg/m³) and Finishers (199 µg/m³), whereas controls averaged only 1.0 µg/m³.

Manganese was markedly elevated in Welders (40.3 µg/m³; SD 34.9) and Finishers (20.8 µg/m³), consistent with Mn-bearing electrode coatings. Silicon was elevated in Welders (92.7 µg/m³), most likely originating from silica-containing electrode flux. Cobalt was highest in Welders (0.80 µg/m³) and Grinders (0.46 µg/m³), consistent with Co-bonded hard-metal tool inserts, and arsenic was highest in Grinders (0.24 µg/m³) and Welders (0.14 µg/m³).

Because site-specific material-specification data were unavailable (see Limitations), the elemental signatures observed across job categories are interpreted in light of material inputs typically reported for comparable metalworking processes. The dominance of Fe reflects the carbon-steel substrate worked throughout the production chain. The much higher CA_Fe in Welders (380 µg/m³) than in Forgers (40.9 µg/m³) is attributable to the arc-welding process, in which electrode and workpiece Fe is volatilized and oxidized at high temperature to form sub-micron Fe-oxide fume (Fe₂O₃/Fe₃O₄)^[47-50]. Forger CA_Fe (40.9 µg/m³), by contrast, reflects mechanical scale generation at the lower temperatures of hammer forging. The elevated K concentrations are most plausibly associated with the charcoal combustion used during forging^[51,52], whereas Si reflects both its role as a minor alloying element in steel and its release during high-temperature processing^[53].

Manganese concentrations were highest in Welders (40.3 µg/m³), consistent with the 0.5%-2.5% Mn content of mild-steel electrodes. In welding fume, Mn occurs predominantly as Mn₃O₄ and MnO, oxidation states that cross the blood–brain barrier more efficiently than others^[54,55]. The mean CA_Mn for Welders (40.3 µg/m³ = 0.040 mg/m³), or 0.030 mg/m³ after 8-h TWA adjustment (CA × ET/8), exceeds the ACGIH Threshold Limit Value for Mn fume (0.02 mg/m³) by ~1.5-fold and is far above the levels (~0.001 mg/m³) that some studies have linked to subclinical neurological effects, indicating severe overexposure^[56-58].

The higher Co levels in Welders (0.8 µg/m³) and Grinders (0.5 µg/m³) most likely arise from the use of cobalt-bonded tungsten carbide (WC–Co) tools. Cobalt exposure during the cutting and grinding of such hard metals is a well-recognized cause of hard-metal lung disease^[59,60], and even low concentrations are of concern given cobalt’s high toxicity and established carcinogenic potential^[61].

Non-carcinogenic risk results are presented in Supplementary Tables 5-7, Table 1 and Figure 3. Manganese was overwhelmingly the dominant HQ contributor, accounting for > 90% of total HI in Welders and Finishers. Mean HQ_Mn ranged from 172 (Welders) to 13.7 (Forgers) across job categories, compared with 0.63 in controls. Cobalt was the second most important driver, with mean HQ_Co = 8.5 (Welders), 4.9 (Grinders), and 4.5 (Finishers). Nickel produced a notable HQ_Ni of 9.1 in Grinders (100% exceeding HQ = 1), reflecting the low RfC for Ni combined with high PM loading and long exposure duration. Arsenic contributed mean HQ_As = 3.5 (Grinders) and 1.9 (Welders), whereas chromium, copper, and zinc contributed negligible HQ values (< 0.01) across all groups.

The HI exceeded the acceptable limit (HI > 1) in all five blacksmith categories [Supplementary Table 6]. Welders recorded the highest mean HI (184 ± 160; 100% exceedance), followed by Finishers (96.4 ± 94.6), Grinders (63.2 ± 65.8), Cutters (21.4 ± 28.2), and Forgers (15.7 ± 25.9; 85.7% exceedance). Controls had a mean HI of 1.9 (SD 1.4; 70% exceeding HI = 1), indicating low-level community exposure to Mn, Ni, and Co. Workers had significantly higher HI than controls (Table 1; Kruskal–Wallis H = 24.32, P < 0.001; Mann–Whitney U = 271, P = 0.0002).

Manganese dominates the non-carcinogenic risk profile, contributing 90%-95% of total HI in Welders and Finishers. The mean HI values of 184 (Welders) and 96.4 (Finishers) correspond to 184-fold and 96-fold exceedances of the acceptable HI = 1 threshold. These are among the highest values reported in the occupational Mn-exposure literature and indicate a severely elevated neurotoxic risk. Early signs of Mn neurotoxicity, including psychomotor slowing, tremor, and mood disturbances^[62-64], have been documented in welders exposed to air Mn concentrations of 0.11-0.46 mg/m^3[62], a range that the mean CA_Mn for Welders in this study (0.04 mg/m³) closely approaches, pointing to a critically elevated risk of manganism.

Cobalt is the second major non-carcinogenic risk driver, with HQ_Co of 8.5 (Welders), 4.9 (Grinders), and 4.5 (Finishers); values well above unity indicate an elevated risk of hard-metal lung disease. The unexpected finding of HQ_Ni = 9.1 with 100% exceedance in Grinders, despite a relatively low absolute CA_Ni (0.60 µg/m³), reflects the combined effect of a very low RfC (0.014 µg/m³; US EPA IRIS) and Ni concentrations elevated relative to other job categories. Because HQ is independent of exposure duration, this hazard is distinct from the duration-sensitive carcinogenic pathway and indicates that even short-term Ni exposure at this concentration carries meaningful non-carcinogenic risk. Nickel compounds are classified as carcinogenic to humans (IARC Group 1)^[21], and occupational exposure has been linked to increased risks of lung and nasal cancers^[65,66]; nickel exposure should therefore be monitored in Grinders as part of occupational-health surveillance.

The 70% HI > 1 exceedance among controls (mean HI = 1.93) suggests a marginal but non-negligible community-level risk from Mn and Co, likely originating from agricultural Mn inputs, road dust, and other ambient sources. Although well below worker levels, this residual community risk should be incorporated into total risk estimates when communicating findings to the affected population. and may warrant independent investigation of ambient Mn and Co sources in the surrounding community.

Carcinogenic risk results are presented in Supplementary Tables 5-7 and Figure 4. Cobalt was the principal ILCR driver across all blacksmith groups, with mean ILCR_Co of 4.8 × 10^-4 (± 7.2 × 10^-4) for Welders, 1.6 × 10^-4 (± 2.3 × 10^-4) for Grinders, and 7.6 × 10^-5 (± 8.1 × 10^-5) for Finishers. Arsenic contributed a minor but non-negligible ILCR in Grinders and Welders, whereas the contributions of Ni and Pb were negligible in all groups.

The upper ILCR threshold (1 × 10^-4) was exceeded by 42.9% of Welders, 66.7% of Grinders, and 50.0% of Finishers [Supplementary Figure 2], while Forgers and Cutters showed no exceedance. Controls exhibited 10% exceedance, consistent with elevated ambient Co in the surrounding community. Workers had significantly higher ILCR_total than controls (Kruskal–Wallis H = 18.07, P = 0.003; Mann–Whitney U = 222, P = 0.026; Table 1).

Mirroring the non-carcinogenic results, cobalt is the principal contributor to carcinogenic risk. Mean ILCR_Co in Welders (4.8 × 10^-4) and Grinders (1.6 × 10^-4) lies well above the 1 × 10^-4 upper acceptable threshold, whereas in Finishers the mean (7.6 × 10^-5) falls just below it. This Co-dominated cancer-risk profile is unusual in the metalworking ILCR literature, which more commonly identifies Cr(VI) and Ni as the principal drivers in formal-sector manufacturing^[7,59,65]. Its prominence here reflects the specific WC–Co hard-metal tooling used in West Java blacksmith production and underscores the value of job-category-specific elemental analysis over generic metalworking-exposure characterization.

Figure 5 and Table 1 present the direct worker-versus-control comparison. Workers had significantly higher total PM (median 1,467 vs. 103 µg/m³; Mann–Whitney U, P = 0.000003), HI (median 40.7 vs. 2.2; P = 0.0002), and ILCR_total (P = 0.026) than controls. Spearman correlation analysis among workers [Figure 6] identified CA_Fe and CA_Mn as the elements most strongly correlated with the aggregate risk indices (CA_Fe–HI, ρ = 0.97; CA_Mn–HI, ρ = 0.98; CA_Fe–ILCR_total, ρ = 0.83; CA_As–ILCR_total, ρ = 0.85; all P < 0.001). Metal-specific scatter analyses of the pooled worker-and-control dataset [Supplementary Figure 3] resolved these relationships at the level of individual hazard and cancer indices, confirming strong positive associations between CA_Fe and HI (ρ = 0.951, P < 0.001), CA_Mn and HQ_Mn (ρ = 0.998, P < 0.001), and CA_As and ILCR_As (ρ = 0.895, P < 0.001). Because Mn and As are major contributors to their respective risk indices, and Fe is a strong correlate of HI despite not entering its calculation, the consistency of these associations across both the worker-only [Figure 6] and pooled worker–control [Supplementary Figure 3] analyses supports a common occupational source for these elements.

The magnitude of the risks documented here makes intervention imperative, and three measures should take priority. First, source-directed local exhaust ventilation (LEV) installed within 0.3-0.5 m of the welding arc and grinding point can be expected to reduce breathing-zone PM and Mn by 70%-95%, based on international LEV performance data^[67-69]; general dilution ventilation is inadequate given the extreme point-source emission rates observed. Second, P100 half-face respirators (≥ 99.97% filtration efficiency) should be provided and mandated for all welding, grinding, and finishing tasks. Third, a periodic biological-monitoring program for blood/urinary Mn and urinary Co should be established for all workers with more than one year of employment, given the established value of these biomarkers for manganism and cobalt-related hard-metal lung disease, respectively.

These recommendations must be viewed within the Indonesian regulatory context. The current NAAQS (55 µg/m³ PM_2.5) applies to ambient air only; no comparable standard yet exists for indoor or occupational environments, particularly for finer fractions such as PM₁. The standard also omits Co from its enforceable parameters and lacks the PM sub-fraction limits found in the US OSHA and EU OEL systems. Extending the regulatory framework to include enforceable workplace exposure limits for Co, Mn, and Ni would bring Indonesian occupational-health standards into closer alignment with international best practice.

Several limitations warrant acknowledgment. First, the cross-sectional design precludes causal inference regarding health outcomes, and concurrent spirometry, neurological assessment, and biological-monitoring data were not collected. Second, the per-category sample sizes are small (Finishers, n = 4; Cutters, n = 3), limiting the statistical power of intra-category comparisons. Third, monitoring was conducted in a single season, so seasonal variation in PM and elemental concentrations cannot be excluded. Fourth, the US EPA framework applies point estimates of IUR and RfC without Monte Carlo uncertainty propagation, which may understate individual risk variability. Fifth, the PNS collected only two sub-micron fractions (PM_0.5 and PM_0.5-1); the full inhalable and respirable occupational fractions were not captured, and comparisons with PM_2.5 standards are therefore contextual only. Sixth, EDXRF measures total elemental concentrations and cannot resolve chemical species or oxidation states, so all risk estimates reflect worst-case assumptions. Seventh, Grinder samples with extremely high PM loadings, filter overloading, particle bounce, and XRF matrix effects cannot be entirely ruled out. Eighth, concurrent ambient PM data were unavailable, preventing precise characterization of the non-occupational background contribution to control-group PM levels. Ninth, source attribution of the elemental profiles to specific material inputs (steel, electrodes, tooling) rests on consistency with the published metalworking literature rather than site-specific data, because safety data sheets and alloy-composition records were not available from the informal-sector workshops studied.