An integration approach for health risk assessment and priority control factor identification of indoor phthalates in a plastic waste recycling plant

Abstract

Phthalates (PAEs) contained in plastic waste can be released to the surrounding environment during recycling and reprocessing. However, contamination profiles and associated occupational health risks within indoor environments have yet to be comprehensively explored. In this study, the occurrence of eight PAEs was investigated in dust and particulate matter collected from a plastic waste recycling plant in South China. The Principal Component Analysis (PCA) and Positive Matrix Factorization (PMF)‌ models were used to identify the sources of dust PAEs, while the Health Risk Assessment (HRA)‌ model with ‌Monte Carlo simulation were applied to assess occupational health risks. As the results, PAEs concentrations in dust and particulate matter within the ranges of 94.27–514 µg·g⁻¹ and 483–1259 ng·m⁻³ inside the plastic waste recycling plant, respectively. DEHP, DBP, and DIBP dominated in both dust and particulate matter. Three sources were identified including building materials (11%), plastic waste feedstock (23%) and plastic waste reprocessing (66%). The HRA model indicated an acceptable carcinogenic risk for occupational workers (adult females: 99.98%>10⁻⁶ ; adult males: 86.39%>10⁻⁶), with negligible non-carcinogenic risk. Compared with adult males, adult females faced higher potential health risks, with ingestion was the dominant exposure route. Plastic waste reprocessing sources and DEHP were recognized as primary control factors, indicating the need for strengthened process management. This study provides a scientific basis for accurately identifying the priority control factors in plastic waste recycling plants and formulating effective pollution prevention and control measures.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-51188-5.

Introduction

As plasticizers and softeners, phthalates (PAEs) are extensively applied in medical devices, toys, cosmetics, automotive components, food packaging, and electronic products^1–4. The annual global PAE consumption is considerable, exceeding 8 million tons⁵. The Handbook of Chemical Economics indicates that PAEs account for approximately two-thirds of the total plasticizer consumption. China is the largest plasticizer market worldwide, with an estimated approximately 38% share of the global market¹. Because PAEs are noncovalently bound to plastics via hydrogen bonding or van der Waals forces, with modestly low binding energies, they can readily enter the environment during plastic production, use, and disposal and are therefore widely present in multiple environmental matrices, including the atmosphere, water, soil, and sediments^6–10. As typical endocrine-disrupting chemicals, PAEs can disturb human reproductive development and nervous system function significantly^11,12. In cohort studies, chronic exposure to PAEs in women is associated with decreased estrogen levels, impaired fertility, and elevated miscarriage rates¹³, and that PAE concentrations in the urine of pregnant women are closely link to pregnancy complications, such as anemia and preeclampsia¹⁴. Animal experiments have also found that di-2-ethylhexyl phthalate (DEHP) exposure cause reproductive toxicity, including malformations of the reproductive tract of male offspring and decreased sperm counts¹⁵. Given their demonstrated risks to the environment and human health, PAEs are listed as priority-control environmental pollutants in numerous countries globally.

In recent decades, the amount of household- and industry-derived plastic waste has increased markedly worldwide, and the associated environmental burden is a major global concern^16,17. Currently, physical recycling technologies are the mainstream strategy for resource recovery of plastic waste¹⁸. However, hazardous chemicals commonly present in plastic waste, such as PAEs, may be released into environmental media during recycling processes, thereby exacerbating secondary pollution risk^19,20. The environmental impacts of municipal solid-waste and electronic-waste recycling have been recognized^6,21–24, whereas attention to plastic waste recycling activities remains relatively limited. In addition, research has primarily focused on soil and sediments^17,25, largely overlooking their associated impacts on dust and air. Only one study has characterized the atmospheric pollution profiles of PAEs emitted from plastic waste recycling plants. Huang et al.²⁶, reported that the concentrations of PAEs released from plastic waste recycling granulation processes were higher than those at a background site (particle-phase: 263.4 vs. 40.9 ng·m⁻³, gas-phase: 938.8 vs. 27.6 ng·m⁻³). However, emission standards for PAEs from plastic waste recycling plants remain absent, and the chronic occupational exposure of workers necessitates comprehensive evaluation of potential health impacts. Concurrently, the potential environmental risks of plastic recycling and reprocessing activities remain unclear because contamination profiles and PAE-related hazards have yet to be comprehensively identified.

Human health risk assessment (HRA) model is one of the principal risk evaluation approaches endorsed by the U.S. Environmental Protection Agency (USEPA). However, existing studies indicate that in conventional health risk assessments, deterministic factors might not reflect actual risk because of variability in concentrations and exposure parameters, potentially resulting in under- or overestimation²⁷. To address the limitation above, the incorporation of probabilistic approaches into traditional deterministic risk assessments has been attempted, with considerable success in capturing inherent uncertainties^28,29. Among the approaches, Monte Carlo simulation (MCS) has attracted significant attention in health risk assessment because it quantifies uncertainty by simulating a large number of variables and thus enhance the reliability of risk evaluation^30–32. For instance, Tong et al.³³, applied MCS to HRA of soil polycyclic aromatic hydrocarbons, revealing that 45% of the total carcinogenic risk (TCR) values exceeded the acceptable threshold (10⁻⁶). In contrast, deterministic assessment yielded a conservative TCR estimate of 1.73 × 10⁻⁷, demonstrating significant underestimation of the actual health risks. Probabilistic health risk assessment thus achieves more accurate characterization of health risks by providing the quantification of uncertainty. Furthermore, health risk assessment can be integrated with source apportionment to form source-oriented risk assessment, which enables more accurate identification of major pollution sources and their associated health risks^34–36. For example, Shen et al.³⁷, combined the positive matrix factorization (PMF) model with HRA to explore the health risks of heavy metals under different pollution sources in urban soil of Shanghai, and they reported that metal product sources was the main cause of non-carcinogenic risk with a contribution rate of 39.71% while the chemical industry and mixed transportation sources was the main cause of carcinogenic risk with a contribution rate of 36.99%.

In summary, while existing studies have established a significant association between plastic waste recycling and PAEs contamination, there remains limited understanding of whether PAEs contamination profiles impact health risks for occupational workers and to what extent such risks are present in the indoor environments of plastic waste recycling plants. This knowledge gap severely constrains the scientific assessment and targeted management of environmental risks associated with plastic recycling processes, emphasizing the necessity of integrated investigations into PAEs contamination profiles, source apportionment, and health risk evaluation within plastic recycling plants. Given this, a source-oriented risk assessment framework can thus be developed by integrating the PMF model, HRA model, and MCS, enabling quantification of source‑specific contributions to health risks and supporting risk‑based management of PAEs pollution. The contents of our study were as follows: (1) to investigate the concentration levels and compositional characteristics of PAEs in dust and total suspended particle (TSP) in a plastic waste recycling plant; (2) to perform source identification of PAEs in the dust using the principal component analysis (PCA) and PMF models; and (3) to evaluate health risks of PAEs for occupational workers employing probabilistic models and to integrate PMF apportionment results to refine source-specific health risk assessment.

Materials and methods

Chemicals and reagents

Eight PAEs, namely dimethyl phthalate (DMP), diethyl phthalate (DEP), diisobutyl phthalate (DIBP), dibutyl phthalate (DBP), butyl benzyl phthalate (BBP), di-2-ethylhexyl phthalate (DEHP), di-n-octyl phthalate (DNOP), and dinonyl phthalate (DNP) were analyzed. Detailed information on the chemicals and reagents is provided in the Text S1 and Table S1.

Study area and sample information

Samples were collected from a plastic waste recycling plant in Shanghai, China. The plant is engaged primarily in plastic waste collection, shredding, granulation, and injection molding, and it processes approximately 70,000 tons of plastic waste annually derived from household and industry. There are five different plastic recycling workshops in the plant: SF1, the polyethylene recycling workshop (granulation and storage); SF2, the injection molding workshop (trial run, producing mainly automotive parts and charging piles); SF3, the ABS recycling workshop (injection molding, new materials); SF4, the polypropylene recycling workshop (shredding and storage); and SF5, the polyethylene recycling workshop (granulation and storage, suspend). Workers in the workshops worked typically 8 h·day⁻¹ and were required to wear protective work uniforms, gloves, and dust masks.

From March 15 to March 18, 2023, 49 samples were collected including 39 dust and 10 total suspended particle (TSP) samples. Dust was sampled from the interior of the workshops (AVG: SF1–SF5), outdoor areas adjacent to the workshops (SFO: N, S, W, E, and NW), and roads outside the plant (R: R1–R3). TSP was sampled from the five different workshops. The sampling site information is shown in Fig. 1, and detailed sampling information is provided in Text S2 and Table S2.

Fig. 1.

Locations and spatial distribution of the sampling sites.

Chemical analysis, quality assurance, and quality control

The PAEs in the dust and TSP were pretreated based on the optimized method described by Li et al.³⁸,, respectively. Target compounds were quantitatively detected using a TQ-8040 NX gas chromatograph–mass spectrometer (Shimadzu, Tokyo, Japan). Detailed information on sample preparation and instrumental parameters is provided in Text S3 and Table S3.

Quantification was performed using the internal standard method with DMP-d4, DBP-d4, and DEHP-d4 serving as the internal standards. The solvent-spiked recoveries of the target PAEs ranged from 75 to 104% and 79–118% for the dust and TSP, respectively. The matrix-spiked recoveries ranged from 77 to 115% and 79–118% for the dust and TSP, respectively. The relative standard deviations were all < 15%, meeting the requirements for environmental analysis. Additional detail information on quality assurance and quality control (QA/QC) are provided in Text S4, Tables S4 and S5.

Positive matrix factorization (PMF) model

The Positive Matrix Factorization (PMF) method first estimates component‑specific errors through weighting, then uses least‑squares optimization to identify major pollution sources and quantify their respective contribution rates^39,40. Detailed descriptions of PMF model are provided in Text S5.

Health risk assessment (HRA) model

The health risk assessment (HRA) model was constructed according to the USEPA Human Health Risk Assessment Guidelines, and the specific formulae are provided in Text S6. To improve the accuracy and reliability of the health risk assessment, MCS was used to address model-related uncertainties and the simulation was performed using the Oracle Crystal Ball (Oracle, Austin, TX, USA) software^28,41. Detailed descriptions of MCS are provided in Text S7, and the exposure parameters and probability distributions for the model inputs are listed in Tables S6-S8.

Results and discussion

Pollution characteristics of PAEs in dust and TSP

Concentration levels

Dust

The concentrations and detection frequencies of eight PAEs in indoor and outdoor dust from the plastic waste recycling plant are illustrated in Fig. 2a. The detection frequency of 100% for all PAEs demonstrated their ubiquitous existence in the plant environment. The dust concentrations of ∑₈PAEs from the recycling workshops ranged from 94.27 to 514 µg·g⁻¹, with a geometric mean of 262 µg·g⁻¹ and a median of 125.92 µg·g^{− 1}. The values were far above the levels observed in outdoor workshop dust (28.21–255 µg·g⁻¹; mean: 61.15 µg·g⁻¹; median: 40.12 µg·g⁻¹) and off-site dust (4.71–29.01 µg·g⁻¹; mean: 17.64 µg·g⁻¹; median: 18.17 µg·g⁻¹). The concentrations showed a pronounced decline from the workshop interiors toward the areas outside the plant, indicating a close association between PAEs and plastic waste recycling activities and suggesting that the activities contributed to the surrounding PAEs contamination. In the present study, indoor dust collected from the recycling workshops contained higher concentrations of PAEs than those reported for residential indoor environment across East China (median: 107.59 µg·g⁻¹)³⁸, further supporting that plastic recycling activities served as major potential emission sources of PAEs. Deng et al.⁶, reported that the concentrations of ∑₉PAEs in indoor dust from the e-waste recycling workshops (median 1400 µg·g⁻¹) were significantly higher than those in local rural homes (median 300 µg·g⁻¹) and urban homes (median 680 µg·g⁻¹), as well (Table S9). In addition, plastic recycling activities have been reported to increase the PAEs concentrations in soils and sediments relative to background levels. For example, the mean concentrations of ∑₂₂PAEs in soils and sediments surrounding a plastic market in Yuyao, China, were 15.1 and 30.7 µg·g⁻¹, respectively¹⁷. Similarly, the mean concentrations of ∑₆PAEs around a plastic solid-waste recycling site in China were 14.4 µg·g⁻¹ in soils and sediments and 31.7 µg·g⁻¹ in sediments²⁵.

Fig. 2.

PAEs concentrations measured in dust and TSP from the plastic waste recycling plant. (a) In indoor and outdoor dust (concentrations are presented as box plots; boxes represent the 25th and 75th percentiles; squares indicate means; horizontal lines indicate medians; whiskers represent the 10th and 90th percentiles; dots denote outliers), (b) Dust (µg·g⁻¹) from different workshops, (c) TSP (ng·m⁻³) from different workshops.

The PAEs concentrations in indoor dust from five recycling workshops are illustrated in Fig. 2b. The average levels of ∑₈PAEs declined in the sequence of: SF5 (417 µg·g⁻¹) > SF3 (411 µg·g⁻¹) > SF1 (263 µg·g⁻¹) > SF2 (121 µg·g⁻¹) > SF4 (97.65 µg·g⁻¹). The trends observed above may be explained by differences in workshop functions and operating conditions. For example, PAEs concentrations varied with both the feedstock type and the specific temperatures applied in the melting and extrusion processes from the plastic wastes recycling granulation processes²⁶. DEHP was the most abundant congener among eight PAEs, with concentrations ranging from 86.31 to 383.55 µg·g⁻¹ (median: 174.85 µg·g⁻¹; mean: 223.68 µg·g⁻¹), followed by DBP (5.52–64.74 µg·g⁻¹; median: 22.79 µg·g⁻¹; mean: 25.64 µg·g⁻¹) and DIBP (2.76–20.56 µg·g⁻¹; median: 4.85 µg·g⁻¹; mean: 7.56 µg·g⁻¹). Generally, DEHP and DBP serve commonly as plasticizers in PVC, cellulose esters, polyvinyl acetate and polyurethane^12,42.

TSP

The PAEs concentrations in the TSP from five recycling workshops are presented in Fig. 2c. All PAEs except DNP were detected in 100% of the samples. For ∑₈PAEs in TSP, the concentrations ranged from 483 to 1,259 ng·m⁻³, with both the geometric mean and median equal to 865 ng·m⁻³. Our results are broadly comparable to, and in certain instances marginally exceed, previously reported values. For example, the PAEs concentrations in the particle-phase were 263.4 and 279.7 ng·m⁻³ inside both plastic waste recycling plants in Xingtang Town, southern China, respectively, which significantly higher than the background concentrations (40.9 ng·m⁻³)²⁶. Similarly, Gu et al.⁴³, reported that the concentrations of ∑₅PAEs in PM_2.5 from the e-waste dismantling area in Taizhou were higher than that at background site both in summer and in winter (summer 216.41 vs. 106.16 ng·m⁻³, winter 326.60 vs. 197.08 ng·m⁻³)(Table S9).

The average levels of ∑₈PAEs declined in the sequence of: SF5 (1,259 ng·m⁻³) > SF1 (1,112 ng·m⁻³) > SF2 (865 ng·m⁻³) > SF3 (605 ng·m⁻³) > SF4 (483 ng·m⁻³). Consistent with the results in dust, DEHP, DBP, and DIBP were the principal PAEs in the indoor TSP from the workshops as well. Notably, DEHP dominated in SF1 (811 ng·m⁻³) and SF5 (477 ng·m⁻³), whereas in SF2, SF3, and SF4, concentrations of DBP (145–411 ng·m⁻³) and DIBP (178–258 ng·m⁻³) were remarkably exceeded those of DEHP (30.9–91 ng·m⁻³). The differences might be attributable to workshop-specific production processes, variations in dust–air partitioning behavior, and interactive effects among emission sources, physicochemical properties, and ambient temperature and humidity, resulting in distinct PAEs emission patterns^26,44,45.

Compositional characteristics

Dust

As illustrated in Fig. 3a, the compositional profiles of PAEs were characterized in both indoor and outdoor dust from the plastic waste recycling plant. DEHP, DBP, and DIBP were the dominant‌ PAEs, with comparable compositions observed in indoor and outdoor dust. Within the workshops, DEHP (85.44%) was the predominant congener, followed by DBP (9.79%) and DIBP (2.89%); together, the three compounds accounted for 98.12% of ∑₈PAEs. Published studies have identified that DEHP, DBP and DIBP as the major PAE congeners in indoor dust across China^12,38. The findings suggest that PAEs can be emitted from various plastic-based materials and that PAEs production and use in China has followed a relatively stable and specific pattern. Similar compositional profiles of PAEs were observed in indoor dust samples from different workshops (Fig. 3b). DEHP accounted for 66.48–92.31% of ∑₈PAEs, followed by DBP (5.55–24.62%) and DIBP (1.09–7.82%).

Fig. 3.

Relative contributions of individual PAEs in dust and TSP from the plastic waste recycling plant. (a) Indoor and outdoor dust, (b) Dust from different workshops, (c) TSP from different workshops.

TSP

The compositional profiles of PAEs in indoor TSP from different workshops are presented in Fig. 3c. Consistent with the dust profiles, DEHP, DBP, and DIBP were the dominant‌ congeners, comprising 34%, 28%, and 26% of ∑₈PAEs, respectively. However, the proportions of DBP and DIBP in TSP were markedly elevated compared with those in dust. The distinct PAEs profiles observed in the air and dust samples can be attributable to diverse inputs from plastic waste recycling activities, as well as the exchange of PAEs between settled dust and airborne particles throughout atmospheric transport⁴⁴. Moreover, PAEs exhibit a higher affinity for adsorption onto smaller particles, a phenomenon that is particularly pronounced in indoor environments⁴⁶. Notably, the compositional profiles varied among workshops. DEHP was predominant in SF1 (72.91%). In SF5, DEHP, DBP, and DIBP accounted for 37.89%, 31.06%, and 24.26%, respectively. DBP and DIBP were the predominant congeners in SF2, SF3, and SF4, with DBP accounting for 47.51%, 30.25%, and 30.02%, respectively, and DIBP accounting for 27.17%, 42.64%, and 36.85%, respectively.

Source apportionment of PAEs in dust

Source apportionment using PCA

In the results of PCA analysis, three principal components were extracted based on the threshold of eigenvalues greater than 1.0 (Fig. S1). PC1 explained 40.2% of the total variance, and showed strong loadings for DEHP (0.839), DEP (0.840), DNOP (0.809), DBP (0.760), and DNP (0.651). PC2 accounted for 25.8% of the total variance and showed strong loadings for DIBP (0.997) and BBP (0.991). PC3 explained 16.4% of the total variance, with DMP exhibiting the highest loading (0.617). High-molecular-weight PAEs, such as DEHP, DNP, and DNOP, are commonly added to plastic products in industrial applications to enhance polymer flexibility and plasticity^47,48. Low-molecular-weight PAEs (e.g., DBP, DEP and DMP) are have been applied ubiquitously in cosmetics and personal care products, and also used as a plasticizer⁸. DIBP and BBP are linked to the application of plasticizers in plastic products and emissions from building materials and furniture⁴⁹.

Source apportionment using PMF

Quantitative source apportionment of PAEs in dust from the study area was conducted using PMF 5.0 to identify sources and their contribution rates. The PMF model resolved three major sources of PAEs in dust from the plastic waste recycling plant. The sources were generally consistent with the PCA results, although some differences were observed. The three factors contributed 11% (Factor 1), 23% (Factor 2), and 66% (Factor 3), respectively (Fig. 4a). There were significant positive correlations among DEHP–DMP (0.74), DIBP–BBP (0.86), DBP–DEP (0.89), and DIBP–DEP (0.83), indicating that the compounds may share the same contamination source. In contrast, DNP exhibited weak correlation with the other PAEs (all < 0.5), suggesting a more independent source (Fig. 4c).

Fig. 4.

Source apportionment of PAEs in dust. (a) Percentage contributions of each factor identified by the PMF model, (b) Distribution of PAEs across the three PMF-resolved factors, (c) Correlations among PAEs derived from Spearman correlation analysis and the PMF model.

Factor 1 was characterized primarily by DNOP (74%), BBP (67%), and DIBP (57%) (Fig. 4b). The compounds have been widely used in adhesives, building sealants, and plastic films and have frequently been detected as additives in construction materials⁵⁰. Therefore, Factor 1 was inferred to represent emissions from building materials. Factor 2 was characterized by DNP (49%) and DEP (22%) (Fig. 4b). DNP has been often used as a plasticizer in various plastic products due to its ability to enhance the flexibility of polymers⁵¹. Meanwhile, DEP, characterized by its low molecular weight, has been applied commonly in personal care products and adhesives, and is also incorporated into plastics as a plasticizer to enhance the flexibility of plastic products^52,53. Therefore, Factor 2 was inferred to be primarily associated with emissions from plastic waste feedstock. Factor 3 was characterized by high contributions of DEHP (74%), DMP (69%), and DBP (51%) (Fig. 4b). DEHP has been extensively employed as a plasticizer in PVC materials and other synthetic polymers⁵⁴. In contrast, DMP and DBP, owing to their relatively low molecular weights, have been used widely in fixatives and adhesives and are common plasticizers in plastic-related industrial activities⁸. Therefore, Factor 3 was defined as emissions associated with plastic waste reprocessing.

Overall, the PCA (Fig. S1), PMF (Fig. 4b), and correlation analysis (Fig. 4c) results were consistent. Based on principal component loadings and factor contributions, Factor 1 corresponded to PC2, whereas Factor 2 could be interpreted a further decomposition of PC1. Factor 3 could be regarded as a combined contribution of PC1 and PC3. In addition, PAEs with high contributions within the same factor exhibited strong correlations, which is in agreement with earlier studies^6,55. PCA and correlation analysis provided additional support for the reliability of the PMF-derived source apportionment.

Human health risk assessment

Concentration-oriented health risk assessment

The non‑carcinogenic and carcinogenic risks for sex‑stratified workers exposed to PAEs in dust and TSP from the plastic waste recycling plant were assessed using the MCS model with 10,000 iterations. For dust, the exposure routes included ingestion, inhalation, and dermal contact, while for TSP via inhalation. While inhalation is a common exposure pathway for both settled dust and airborne TSP, the former quantifies the contribution from surface dust resuspension, and the latter quantifies the background burden of airborne particulates. The cumulative probability distributions are shown in Fig. 5. The mean hazard index (HI) values were 3.71 × 10⁻₂ for males and 4.20 × 10⁻₂ for females, indicating negligible non-carcinogenic risk (Fig. 5a). Meanwhile, the cumulative non-carcinogenic risk in females was greater than in males. The difference might be attributable to lower average body weight in females, sex-related differences in metabolic characteristics, and physiological variation, which could increase sensitivity to PAEs toxicity⁵⁶. Based on the cumulative probability distributions of the hazard quotient (HQ), the average HQ values ranking for both males and females was as follows: DEHP > DBP > DNOP > BBP > DEP > DMP. All HQ values were < 1, indicating negligible non-carcinogenic risk (Fig. 5c and d). The average HQ of DEHP exceeded those of the other five PAEs by 1–4 orders of magnitude, underscoring its role as the dominant contributor of non‑carcinogenic risk. DEHP exhibits high exposure potential, bioaccumulation, and endocrine-disrupting effects⁵⁷.

Fig. 5.

Probability distributions of health risks for occupational workers. (a) Hazard index (HI) and (b) Total carcinogenic risk (TCR), (c & d) Hazard quotient (HQ) and (e & f) Carcinogenic risk (CR) of individual PAEs, (g & h) Hazard index (HI) and (i & j) Total carcinogenic risk (TCR) for different sources.

An acceptable carcinogenic risk posed by PAEs was indicated by the cumulative probability distributions of total carcinogenic risk (TCR) (1 × 10⁻₆ ≤ TCR ≤ 1 × 10⁻₄) (Fig. 5b). The mean TCR values for males and females were 1.29 × 10⁻⁶ and 2.08 × 10⁻⁶, respectively, corresponding to approximately 1.29 and 2.08 times the threshold value (1 × 10⁻⁶). In addition, the cumulative probabilities that TCR fell within 1 × 10⁻⁶–1 × 10⁻⁴ were 86.39% for males and 99.98% for females, indicating that the carcinogenic risk for occupational workers remained within an acceptable range, with females exhibiting a higher risk than males. As shown in Fig. 5e and f, DEHP was also the leading driver to carcinogenic risk. The average carcinogenic risk (CR) values of DEHP exceeded 1 × 10⁻⁶, with probabilities of 86.34% for males and 99.98% for females, suggesting that the associated carcinogenic risk fell within acceptable limits. Evidence from human and rodent studies indicates that DEHP can induce carcinogenesis through multiple molecular signaling pathways, including DNA damage⁵⁸. In addition, the carcinogenic risk from BBP was negligible for both males and females as its average CR values were below 1 × 10⁻⁶.

Sensitivity analysis were carried out for each model parameter to assess their influence on the probabilistic risk estimates (Table S10). Then, the five most influential parameters were identified: DEHP concentration in dust (DEHP_dust), ingestion rate (R_ing), inhalation rate (R_inh), body weight (BW), and exposure frequency (EF). The sensitivity plots for health risks in adult males and adult females are shown in Fig. S2. DEHP_dust emerged as the most dominant factor affecting both non‑carcinogenic and carcinogenic risks. BW ranked second in influence and was negatively associated with risk, and adult females were more sensitive than adult males. The EF also positively contributed to health risks. Health risks were more strongly associated with R_ing than R_inh, which reflected the dominant exposure route was ingestion^44,59.

Therefore, intensive industrial activities in plastic waste recycling plants exacerbate PAEs pollution substantially and posed notable health risks to occupational workers. Among PAEs, DEHP represented the major contribution to both non-carcinogenic and carcinogenic risks, with more pronounced effects via the ingestion route and higher estimated risks in females than in males.

Source-oriented health risk assessment

In this study, health risk assessment was integrated with PMF-based source apportionment to link source identification and adverse health consequences (Text S8). Source-specific concentrations of individual PAEs were estimated according to the respective source contributions. In addition, the probability distributions of non-carcinogenic and carcinogenic risks for sex‑stratified occupational workers were simulated to quantify the contributions of each pollution source to the PAEs-related health risks.

The effects of different sources on the non-carcinogenic risk showed consistent patterns in both sexes (Fig. 5g and h). The average HI values declined in the order of: Factor 3 > Factor 2 > Factor 1, indicating that Factor 3 (plastic waste reprocessing) acted as the primary driver in non-carcinogenic risk. However, the average HI values for all three factors were < 1 for both sexes, suggesting negligible non-carcinogenic risk. Similarly, the source‑specific contributions to carcinogenic risk followed a consistent declining pattern (Fig. 5i and j), and Factor 3 exhibited the greatest carcinogenic risk. For females, the average CR exceeded the 1 × 10⁻⁶ threshold, and the probability that carcinogenic risk fell within the acceptable range was 98.17%. In contrast, for males, the average CR did not exceed 1 × 10⁻⁶, and the probability that carcinogenic risk fell within the acceptable range was 37.83%, indicating substantially lower carcinogenic risk for males than for females. The carcinogenic risk associated with Factor 3 originated primarily from DEHP, DMP, and DBP, which showed high contributions to this factor (74%, 69%, and 51%, respectively).

Source-oriented risk assessment helps to quantify the relative contributions of individual pollution sources to the overall health risks³⁶. Overall, plastic waste reprocessing (Factor 3) was identified as the priority control pollution source, accounting for 73.65% of the non-carcinogenic risk and 74.29% of the carcinogenic risk for occupational workers (Fig. 6). From a pollutant perspective, DEHP should be prioritized for monitoring and control because it contributed substantially to human health risks in the present study (Fig. 6). Therefore, enhancing environmental monitoring during plastic recycling, particularly DEHP, is critical for managing PAEs-related risks and protecting occupational workers and the surrounding ecological environment.

Fig. 6.

Relationships among PAEs, sources, and health risks in dust from the plastic recycling waste plant.

Deterministic health risk assessment

The deterministic risk assessment results for sex‑stratified workers are shown in Fig. S3, Tables S11 and S12. The total daily average exposure doses of PAEs for adult males and adult females were 8.28 × 10⁻⁴ and 9.37 × 10⁻⁴ mg·kg⁻¹·day⁻¹, respectively. In addition, similar studies of indoor dust from multiple regions in China reported total adult PAEs intakes of 7.23 × 10⁻⁵and 1.26 × 10⁻⁴mg·kg⁻¹·day⁻¹, respectively, which were far below the exposure levels calculated in the present study^12,38. It was because that these studies considered only dust exposure but not air and plastic waste recycling activities exacerbated PAEs pollution in dust. For non-carcinogenic risk (Fig. S3a and b), the deterministic and probabilistic assessments yielded consistent results. The HI values derived from both approaches were < 1, suggesting negligible non‑carcinogenic risk. For carcinogenic risk (Fig. S3c and d), the deterministic method yielded lower risk estimates than the probabilistic approach and only adult females in SF5 had a CR exceeding 1 × 10⁻⁶. In contrast, the probabilistic assessment indicated that the cumulative probability of exceeding 1 × 10⁻⁶ was 86.39% for adult males and 99.98% for adult females. The discrepancy likely arises because deterministic assessments use fixed parameter values to generate single-point estimates, which may underestimate or overestimate risk and cannot account fully for the variability and uncertainty captured by probabilistic risk assessment^60,61. As a probabilistic method, Monte Carlo simulation can quantify the relative contribution of each pollution source to the overall health risk by incorporating the stochastic variations of multiple input parameters, thus improving the accuracy of risk assessment³⁰.

Conclusion

In summary, this study provided a systematic investigation and elucidation of the pollution profile, sources, and occupational health risks of PAEs in plastic waste recycling plants. The results demonstrated that PAEs were pervasive in the indoor environment of the recycling plant, with concentrations in dust and particulate matter ranging from 94.27 to 514 µg·g⁻¹ and 483–1259 ng·m⁻³, respectively. SF5 workshop exhibited the highest PAEs concentrations. DEHP, DBP, and DIBP were identified as the most prevalent PAEs in both dust and particulate matter. Combined with PCA and PMF models source apportionment analysis, three sources were identified in dust, including building materials (11%), plastic waste feedstock (23%), and plastic waste reprocessing (66%). Probabilistic risk assessment demonstrated that the non-carcinogenic risk for occupational workers was negligible, while the carcinogenic risk was acceptable (adult males: 86.39% > 10⁻⁶; adult females: 99.98% > 10⁻⁶). Compared with adult males, adult females faced substantially elevated potential health risks, with ingestion identified as the primary exposure route. By integrating sensitivity analysis into the human health risk assessment, DEHP in dust emerged as the most sensitive factor. Source-oriented analysis suggested that plastic waste reprocessing and DEHP were the prioritized control factors for PAEs health risks in the indoor environments of the recycling plant. This study could provide theoretical support for effectively managing and mitigating indoor PAEs pollution risks in the plastic waste recycling plant.

Due to the constraints in on-site sampling, the sample size was relatively small in this study, which to some extent compromised the spatiotemporal representativeness of the findings. Additionally, the traditional source apportionment methods employed in this study such as PCA and PMF, are highly dependent on the physicochemical properties of pollutants and require complicated component analysis experiments, resulting in high manual involvement and substantial research costs. In future research, the sample size will be expanded with more samples collected from multiple sites and seasons to improve the representativeness and reliability of data. Furthermore, machine learning algorithms can be integrated into source apportionment. With strong data-fitting capability, machine learning exhibits outstanding advantages in solving complex data patterns. Compared with traditional numerical models, machine learning can significantly improve computational efficiency and enable automated, rapid, and high-precision identification of pollution sources, thereby providing more efficient technical support for the source tracing and control of PAEs pollution.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Mengyu Liu: Writing-original draft, Formal analysis, Data curation, Methodology, Visualization. Jing Zhao: Writing-review & editing, Supervision, Conceptualization, Resources. Mengfei Yuan: Formal analysis, Data curation, Methodology, Validation. Jianfeng Bai: Funding acquisition, Resources. Xiaolong Song: Funding acquisition, Resources. Ting Xu: Conceptualization, Writing-review & editing. Daqiang Yin: Supervision, Resources. All authors reviewed the manuscript.

Funding

This work was funded by the National Natural Science Foundation of China (42571076 and 42471331).

Data availability

Data will be made available on request.

Declarations

Competing interests

The authors declare no competing interests.