Associations of occupational heat exposure with dyslipidemia among petrochemical workers: insights from a 9-year longitudinal study

Abstract

Objective

This study aims to explore the associations between occupational heat exposure and dyslipidemia among petrochemical workers, to evaluate associations in two-exposure models that include other occupational hazards, and to examine subgroup-specific patterns.

Methods

A total of 30,847 occupational health examination records were collected from two petrochemical plants in Fujian Province between 2013 and 2021. The dataset included occupational exposure information and serum lipid levels (total cholesterol [TC], low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein cholesterol [HDL-C], triglycerides [TG]). Generalized estimating equations (GEE) models were used to analyze the associations of occupational heat exposure with dyslipidemia and its subtypes. Sensitivity analyses and lagged exposure analyses were further performed to assess the potential influence of the healthy worker effect.

Results

In the single-exposure models, most occupational hazards showed significant inverse associations with dyslipidemia. In the two-exposure models, heat exposure was positively associated with high TC after additionally accounting for gasoline (incidence rate ratio, IRR = 1.267, 95% CI: 1.117–1.437) or H₂S (IRR = 1.324, 95% CI: 1.166–1.505). In stratified analyses, these positive associations were more evident among females (heat-gasoline model: IRR = 2.240, 95% CI: 1.639–3.062; heat-H₂S model: IRR = 2.347, 95% CI: 1.736–3.173) and workers aged ≥ 35 years (heat-gasoline model: IRR = 1.317, 95% CI: 1.158–1.497; heat-H₂S model: IRR = 1.385, 95% CI: 1.215–1.578). Sensitivity analyses showed that the inverse association with high TG was more pronounced among workers with ≥ 5 years of employment (IRR = 0.82, 95% CI: 0.75–0.89) than among those with < 5 years (IRR = 0.91, 95% CI: 0.84–0.98). In lagged analyses, the inverse association between heat exposure and high TC was attenuated and no longer significant (IRR = 0.945, 95% CI: 0.817–1.092), whereas the inverse association with high TG remained significant (IRR = 0.883, 95% CI: 0.818–0.954).

Conclusion

The inverse association between occupational heat exposure and dyslipidemia in the single-exposure models may be partly attributable to the healthy worker effect. In two-exposure models additionally accounting for gasoline or H₂S, heat exposure was positively associated with elevated TC, whereas subgroup-specific and lagged analyses indicated that these associations should be interpreted cautiously as exploratory.

Background

The petrochemical industry converts raw materials from oil refining and gas processing into a wide range of products, including gasoline, diesel fuel, asphalt, synthetic fibers, rubber and various chemical intermediates [1]. Its complex production processes involve multiple occupational hazards, such as heat, noise, gasoline, hydrogen sulfide (H₂S), carbon monoxide (CO), sulfur dioxide (SO₂), and benzene. Long-term, low-dose exposure to these hazards may adversely affect workers’ health, safety, and well-being [2–4]. Among these, heat exposure is one of the most critical occupational risks for petrochemical workers, arising from both process-generated heat and outdoor high-temperature environments [5, 6]. Furthermore, the use of personal protective equipment (PPE) may impede heat dissipation and increase the risk of overheating and thermal strain [7]. With the projected increase in the frequency and intensity of extreme heat events, occupational heat exposure poses growing challenges to workers’ health and safety, particularly among those performing physically demanding tasks outdoors.

A growing body of evidence indicates that chronic occupational heat exposure significantly increases the risk of cardiovascular diseases (CVDs) [8–11], although the underlying biological mechanisms remain incompletely understood. Dyslipidemia is a well-established risk factor in the onset and progression of cardiovascular disease [12]. It is typically characterized by elevated total cholesterol (TC), triglycerides (TG), and low-density lipoprotein cholesterol (LDL-C), together with reduced high-density lipoprotein cholesterol (HDL-C) [13]. Heat exposure may influence these lipid parameters. For example, Halonen et al. reported that higher ambient temperatures were associated with increased LDL-C and decreased HDL-C levels, suggesting that temperature-related alterations in serum lipids could be a potential mechanism underlying heat-related CVDs [14]. However, the relationship between occupational heat exposure and serum lipid profiles requires further investigation. Evidence from occupational studies has shown that male manufacturing workers exposed to high temperatures had a higher risk of dyslipidemia than unexposed workers [15]. Similarly, a high prevalence of dyslipidemia (41.4%) was reported among workers in a large petrochemical plant in India [16]. Together, these findings suggest that workplace heat exposure may contribute to metabolic disturbances, but the evidence remains insufficient, particularly in petrochemical settings.

In addition to heat, other petrochemical hazards in the workplace may also influence lipid metabolism. Crude oil–derived chemicals can disrupt lipid metabolism. For example, polycyclic aromatic hydrocarbons (PAHs) and other petroleum-based compounds have been shown to induce hepatic lipid metabolic disorders and impair lipid storage mechanisms [17]. Epidemiological studies have also demonstrated associations between exposure to benzene, toluene, ethylbenzene, and xylene (BTEX) and alterations in lipid profiles [18]. In petrochemical operations, workers are usually exposed to multiple occupational hazards, and such combined exposures may exert synergistic or additive adverse effects on lipid metabolism. For instance, Vangelova et al. [19] found a significantly higher incidence of dyslipidemia among middle-aged and older workers exposed to both heat and noise. Similarly, occupational exposure to xylene, organic solvents, and silica has been identified as a potential risk factor for CVDs in petrochemical workers [20]. Experimental studies have also reported combined effects of heat and environmental pollutants in animal models. Therefore, evaluating the health effects of a single occupational hazard without accounting for potential interactions with co-exposures may lead to incomplete or potentially biased risk estimates [21–24].

This study aimed to investigate the association between occupational heat exposure and dyslipidemia among petrochemical workers, and to further examine its associations in two-exposure models with other occupational hazards. The findings may help improve understanding of heat-related metabolic risk in petrochemical workers, facilitate the identification of vulnerable subgroups, and provide insights for the prevention and early control of dyslipidemia and related CVDs in high-risk occupational populations.

Methods

Study population

According to the requirements of the “Law of the People’s Republic of China on the Prevention and Control of Occupational Diseases” and the “Technical Specifications for Occupational Health Surveillance” (GBZ188-2014), employers are required to organize regular occupational health examinations for workers exposed to occupational hazards and to establish occupational health files [25]. The data used in this study were sourced from two petrochemical companies in Fujian Province, China. During the period from 2013 to 2021, workers underwent routine occupational health examinations at the Minnan Branch of the First Affiliated Hospital of Fujian Medical University, the only qualified local hospital authorized to conduct such examinations.

The occupational health examination items mainly included blood biochemistry, routine blood tests, routine urine tests, upper abdominal color ultrasound, and the collection of occupational and personal information such as demographic characteristics, lifestyle factors, dietary habits, and occupational exposure history. Occupational physicians conducted relevant examinations for petrochemical workers and assisted the companies in establishing electronic occupational health records.

A total of 32,544 occupational health examination records were initially obtained. Workers were included in the analysis if they met the following criteria: (1) aged 18–65 years; and (2) had complete occupational exposure documentation. The following workers were excluded: (1) those with missing key occupational exposure or outcome data; and (2) those with a history of severe cardiovascular disease or malignant tumors. For participants who underwent multiple examinations within the same calendar year, duplicate records were merged and only the most recent examination was retained to ensure data consistency. After applying the inclusion and exclusion criteria and removing duplicate records, 1,697 records were excluded. Ultimately, 6,911 workers were included in the final analysis, contributing 30,847 examination records.

Workers were required to fast from 20:00 on the night before the health examination, and fasting venous blood was collected between 8:00 and 10:00 on the examination day. Serum lipid levels, including TC, TG, HDL-C, and LDL-C, were measured using a fully automated biochemical analyzer (Cobas^®8000 composed of Cobas C701 module and Cobas ISE module). This study was approved by the Ethics Committee of Fujian Medical University (Fujian Ethics Examination No. 2022 − 111).

Definitions of exposure variables and health outcomes

In this study, heat exposure was treated as a categorical occupational hazard recorded in the occupational health examination files. Exposure status was determined from occupational health surveillance records and workplace hazard identification records, rather than from individual-level continuous measurements collected at each examination. The classification thresholds were defined in accordance with GBZ 2.2–2007, Measurement of Physical Factors in the Workplace, Part 7. High-temperature work is defined as an operation in which the average wet-bulb globe temperature (WBGT) index at the worksite is ≥ 25 °C during production activities [26].

Occupational exposure history also included information such as length of service and types of occupational hazards (e.g., benzene, methanol, gasoline, acid anhydrides, carbon monoxide [CO], hydrogen sulfide [H₂S], ammonia [NH₃], and noise). Routine monitoring of these occupational hazard factors was conducted by the Fujian Center for Prevention and Control of Occupational Diseases and Chemical Poisoning, which classified and reported exposures according to the National Occupational Hazard Detection Criteria. Therefore, exposure classification in this study reflected hazard identification at the occupational surveillance level rather than repeated quantitative assessment at the individual level.

Blood lipid abnormalities were defined according to the Guidelines for the Prevention and Treatment of Dyslipidemia in Chinese Adults (Revised 2016): TC ≥ 6.2 mmol/L, TG ≥ 2.3 mmol/L, LDL-C ≥ 4.1 mmol/L, HDL-C < 1.0 mmol/L. Dyslipidemia was defined as the presence of at least one abnormal lipid parameter [27]. Accordingly, participants with hypercholesterolemia, hypertriglyceridemia, hypo-high-density lipoproteinemia, and hyper-low-density lipoproteinemia, or any combination of these conditions were classified as dyslipidemia. Regular smoking was defined as smoking ≥ 1 cigarette per day for 6 months during the previous year, or ≥ 7 cigarettes per week. Occasional smoking was defined as smoking less frequently than regular smoking but not meeting the criteria for non-smoking. Alcohol consumption was defined as alcohol intake at least twice per week, with consumption exceeding 50 g per occasion for more than 6 months.

Statistical analysis

Continuous variables are presented as mean ± standard deviation (SD), and categorical variables as frequencies and percentages (%). Given the longitudinal nature of the health examination data, with repeated measurements clustered within individuals, we employed generalized estimating equations (GEE) models to account for within-subject correlation [28]. Although the primary outcomes (dyslipidemia and its subtypes) are binary, we specified a negative binomial distribution with a log link function rather than a binomial distribution with a logit link. This choice was motivated by three considerations. First, we aimed to estimate incidence rate ratios (IRRs) directly, as odds ratios overestimate relative risks for common outcomes (prevalence > 10%) [29]. Second, preliminary analyses revealed overdispersion (variance > mean), which violates Poisson model assumptions. The negative binomial distribution accounts for such overdispersion by incorporating an extra dispersion parameter [30]. Third, the log link enables direct interpretation of IRRs, representing the relative increase in dyslipidemia rate per examination cycle associated with exposure [31]. Although log-binomial models were theoretically appropriate for binary outcomes when estimating relative risks, they often fail to converge with multiple covariates and clustered data [32, 33]. We specified a first-order autoregressive (AR1) working correlation structure, assuming that measurements taken closer in time are more highly correlated [34]. Robust (Huber–White) standard errors were used to ensure valid inference even if the working correlation structure was misspecified [35].

In the primary analyses, the associations of occupational heat exposure with dyslipidemia and its subtypes were estimated after adjustment for age, gender, BMI, smoking, and alcohol consumption. To examine whether the association between occupational heat exposure and lipid abnormalities differed according to co-exposure to other workplace hazards, we fitted two-exposure GEE models including heat exposure, the corresponding chemical exposure, and a cross-product term between them. Given the log link used in the models, this analysis evaluated interaction on the multiplicative scale.

To evaluate the potential influence of healthy worker effect, three sensitivity analyses were conducted: (1) stratification by employment duration (< 5 vs. ≥5 years), (2) restriction to participants with normal lipid levels at baseline, and (3) exclusion of workers with fewer than 2 years of follow-up. In addition, lagged exposure analyses were performed to further assess whether the observed inverse associations might reflect contemporaneous job assignment or early-career selection. Prior-year exposure status was used to predict lipid abnormalities at the subsequent annual examination among workers with consecutive annual follow-up. Lagged single-exposure models were used to estimate the associations of prior-year heat exposure with high TC, high TG, high LDL-C, and low HDL-C. Lagged two-exposure models for high TC were further fitted by simultaneously including prior-year heat exposure and prior-year gasoline exposure, or prior-year heat exposure and prior-year H₂S exposure, as separate covariates. All lagged models were adjusted for prior-year age, prior-year BMI, prior-year smoking status, prior-year drinking status, and examination year. All statistical analyses were performed using StataSE 16 and SAS 9.4. A two-tailed P < 0.05 was considered statistically significant.

Results

Baseline characteristics of the study population

Table 1 presents the sociodemographic characteristics and occupational hazard exposure profiles of workers with and without occupational heat exposure. Significant differences were observed in gender, age group, and smoking status between the two groups (all P < 0.05), while no significant differences were found in BMI and alcohol consumption (P = 0.961 and P = 0.187, respectively). Workers exposed to occupational heat had a significantly higher proportion of co-exposure to gasoline, nitrogen oxides, sulfur dioxide, hydrogen sulfide, benzene, methanol, and carbon monoxide compared to those without heat exposure (all P < 0.001).

Table 1.

Baseline characteristics of petrochemical workers stratified by occupational heat exposure

Variable	Heat exposure			Non-heat exposure
	Overall	Between	Within	Overall	Between	Within	P-value
	n (%)	n (%)	%	n (%)	n (%)	%
Gender
Male	10,090 (85.0)	4508 (83.6)	100	14,130 (74.5)	4273 (78.5)	100	< 0.001
Female	1782 (15.0)	882 (16.4)	100	4845 (25.5)	1172 (21.5)	100
Age group (years)
≦24	2351 (19.8)	1547 (28.7)	85.2	1354 (7.1)	914 (16.8)	78.2	< 0.001
25–34	3865 (32.6)	2021 (37.5)	86.3	4127 (21.8)	1921 (35.3)	86.6
35–54	5267 (44.4)	2233 (41.4)	96.2	12,256 (64.6)	2945 (54.1)	93.7
≥55	387 (3.3)	205 (3.8)	83.9	1231 (6.5)	407 (7.5)	74.6
BMI (kg/m2)
<24	6451 (54.3)	3251 (60.3)	93.0	10,316 (54.4)	3341 (61.4)	87.9	0.961
≥24	5421 (45.7)	2588 (48.0)	91.4	8659 (45.6)	2919 (53.6)	85.9
Smoking
No	6986 (67.0)	3097 (71.2)	96.4	10,548 (66.0)	2927 (69.7)	96.0	0.008
Occasionally	1447 (13.9)	798 (18.3)	74.5	2132 (13.3)	753 (17.9)	73.7
Heavy	2002 (19.2)	883 (20.3)	87.6	3312 (20.7)	960 (22.9)	86.7
Drinking
No	5220 (50.6)	2498 (58.2)	90.0	7885 (49.8)	2339 (56.5)	88.0	0.187
Yes	5091 (49.4)	2256 (52.6)	90.6	7951 (50.2)	2301 (55.6)	90.5
Gasoline
Yes	7236 (61.0)	3666 (68.0)	92.2	5078 (26.8)	2426 (44.6)	77.7	< 0.001
No	4636 (39.1)	2334 (43.3)	86.1	13,897 (73.2)	4024 (73.9)	88.5
Oxynitrides
Yes	1516 (12.8)	1128 (20.9)	72.6	620 (3.3)	526 (9.7)	63.1	< 0.001
No	10,356 (87.2)	4790 (88.9)	95.4	18,355 (96.7)	5227 (96.0)	97.8
SO2
Yes	1313 (11.1)	797 (14.8)	79.2	647 (3.4)	476 (8.7)	67.1	< 0.001
No	10,559 (88.9)	4886 (90.7)	97.4	18,328 (96.6)	5220 (95.9)	98.2
H2S
Yes	6713 (56.5)	3688 (68.4)	89.8	5248 (27.7)	2600 (47.8)	67.9	< 0.001
No	5159 (43.5)	2426 (45.0)	85.7	13,727 (72.3)	4326 (79.5)	85.1
Benzene
Yes	3107 (26.2)	2093 (38.8)	81.4	3208 (16.9)	1579 (29.0)	71.6	< 0.001
No	8765 (73.8)	3996 (74.1)	92.3	15,767 (83.1)	4616 (84.8)	93.5
Methanol
Yes	2336 (19.7)	1502 (27.9)	84.0	1459 (7.7)	664 (12.2)	70.3	< 0.001
No	9536 (80.3)	4321 (80.2)	95.6	17,516 (92.3)	5129 (94.2)	97.1
CO
Yes	3967 (33.4)	2485 (46.1)	81.2	2225 (11.7)	1392 (25.6)	65.7	< 0.001
No	7905 (66.6)	3756 (69.7)	89.8	16,750 (88.3)	4824 (88.6)	93.9
Total	11,872 (38.5)	5390 (78.0)	59.8	18,975 (61.5)	5445 (78.8)	67.8

Data are presented as n (%). Overall n (%) refers to the total number of occupational health examination records (person-examinations) in each category, with percentages calculated using the total number of records within each exposure group as the denominator. Between n (%) refers to the number of unique individuals contributing to those records, with percentages calculated using the total number of unique individuals within each exposure group as the denominator. Within % represents the proportion of individuals in each category whose all examination records were consistently classified within that same category across the study period, indicating the stability of the characteristic over time within individuals

Prevalence of dyslipidemia by occupational heat exposure

Table 2 shows the prevalence of four types of dyslipidemia (hypercholesterolemia [high TC], hypertriglyceridemia [high TG], hyper-low-density lipoproteinemia [high LDL-C], and hypo-high-density lipoproteinemia [low HDL-C]) among workers with and without heat exposure. The overall prevalence of high TC, high TG, high LDL-C, and low HDL-C in the heat exposure group was 3.7%, 10.6%, 0.2%, and 6.3%, respectively. In contrast, the non-heat exposure group had significantly higher prevalence of all four dyslipidemia subtypes: high TC (6.0%), high TG (19.3%), high LDL-C (0.5%), and low HDL-C (9.4%). In subgroup analyses, the prevalence of high TC (8.3% vs. 5.1%), high TG (19.9% vs. 12.4%), high LDL-C (0.9% vs. 0.4%), and low HDL-C (12.8% vs. 9.2%) was also higher in the non-heat exposure group (all P < 0.001).

Table 2.

Lipid disorders by heat exposure among petrochemical workers

Variable	Heat exposure			Non-heat exposure
	Overall n (%)	Between n (%)	Within %	Overall n (%)	Between n (%)	Within %	P-value
High TC
Yes	444 (3.7)	274 (5.1)	66.3	1135 (6.0)	453 (8.3)	45.8	< 0.001
No	11,428 (96.3)	5287 (98.1)	98.5	17,840 (94.0)	5377 (98.8)	97.4
High TG
Yes	1258 (10.6)	667 (12.4)	72.8	3662 (19.3)	1086 (19.9)	59.7	< 0.001
No	10,614 (89.4)	5068 (94.0)	96.8	15,313 (80.7)	5139 (94.3)	93.3
High LDL-C
Yes	23 (0.2)	19 (0.4)	52.6	87 (0.5)	51 (0.9)	28.9	< 0.001
No	11,849 (99.8)	5386 (99.9)	99.9	18,888 (99.5)	5443 (99.9)	99.8
Low HDL-C
Yes	752 (6.3)	494 (9.2)	62.9	1775 (9.4)	697 (12.8)	46.2	< 0.001
No	11,120 (93.7)	5228 (97.0)	97.2	17,200 (90.6)	5350 (98.3)	95.8

Single-exposure model: associations between individual occupational hazards and dyslipidemia

Figure 1 summarizes the associations between individual occupational hazards and dyslipidemia (adjusted for age, gender, BMI, smoking, and alcohol consumption). For TC, most occupational hazards showed inverse associations except heat, benzene, and CO. All hazards were negatively associated with TG (IRRs < 1.0, P < 0.05). For LDL-C, all exposures except CO demonstrated inverse associations. For HDL-C, negative associations were observed for all hazards except hydrogen sulfide (H₂S) and benzene (IRRs < 1.0, P < 0.05).

Fig. 1.

Associations of individual occupational hazards with blood lipid levels in petrochemical workers. Note: All models were adjusted for age, gender, body mass index, smoking, and alcohol drinking

Two-exposure models: interaction analyses on the multiplicative scale

Table 3 presents the results of the two-exposure GEE models, in which occupational heat exposure, one co-occurring chemical hazard, and their cross-product term were included simultaneously. Patterns consistent with possible effect modification on the multiplicative scale were observed for high TC in the models involving gasoline and H₂S. Specifically, in the heat-gasoline model, heat exposure was positively associated with high TC (IRR = 1.267, 95% CI: 1.117–1.437), and a similar pattern was observed in the heat-H₂S model (IRR = 1.324, 95% CI: 1.166–1.505). No similarly consistent pattern was observed for TG, LDL-C, or HDL-C, or for the other chemical hazards. Given the number of interaction tests conducted across multiple hazards and lipid markers without formal correction for multiple comparisons, these statistically significant associations for total cholesterol should be interpreted cautiously in light of the potential for type I error.

Table 3.

Associations of heat exposure and co-exposures with blood lipid levels in two-exposure GEE models

Variable	IRR (95% CI)
	TC	TG	LDL-C	HDL-C
Heat + Gasoline
Heat	1.267 (1.117–1.437)	0.847 (0.788–0.911)	1.077 (0.976–1.187)	0.997 (0.990–1.003)
Gasoline	0.644 (0.529–0.785)	0.583 (0.527–0.645)	0.667 (0.578–0.769)	0.997 (0.989–1.006)
Heat + Oxynitrides
Heat	1.073 (0.962–1.198)	0.800 (0.753–0.851)	0.897 (0.824–0.977)	0.993 (0.989–0.998)
Oxynitrides	0.807 (0.484–1.343)	0.798 (0.637-1.000)	0.414 (0.214–0.803)	0.983 (0.974–0.991)
Heat + SO2
Heat	1.077 (0.966-1.200)	0.812 (0.764–0.863)	0.918 (0.844–0.998)	0.993 (0.989–0.998)
SO2	1.002 (0.649–1.545)	1.041 (0.848–1.278)	1.307 (0.984–1.734)	0.988 (0.977–0.999)
Heat + H2S
Heat	1.324 (1.166–1.505)	0.865 (0.808–0.926)	1.076 (0.974–1.188)	0.994 (0.988-1.000)
H2S	0.923 (0.792–1.076)	0.836 (0.775–0.901)	0.945 (0.842–1.060)	0.997 (0.989–1.004)
Heat + Benzene
Heat	1.059 (0.948–1.182)	0.782 (0.734–0.833)	0.863 (0.790–0.942)	0.993 (0.988–0.999)
Benzene	1.004 (0.823–1.225)	0.861 (0.774–0.957)	0.777 (0.648–0.932)	0.994 (0.985–1.003)
Heat + Methanol
Heat	1.092 (0.979–1.217)	0.781 (0.733–0.831)	0.906 (0.831–0.986)	0.994 (0.989–0.998)
Methanol	0.951 (0.709–1.277)	0.723 (0.609–0.857)	0.975 (0.772–1.232)	0.992 (0.981–1.003)
Heat + CO
Heat	1.119 (0.998–1.226)	0.798 (0.749–0.849)	0.920 (0.842–1.006)	0.992 (0.987–0.998)
CO	1.126 (0.924–1.372)	0.888 (0.806–0.979)	1.068 (0.907–1.259)	0.989 (0.981–0.998)

Stratified analysis: subgroup-specific associations across demographic and occupational characteristics

Figure 2 shows the results of stratified analyses by gender, age, BMI, smoking status, and drinking status. In the two-exposure models including both heat exposure and gasoline, the positive association between heat exposure and high TC appeared more pronounced among females (IRR = 2.240, 95% CI: 1.639–3.062), individuals aged ≥ 35 years (IRR = 1.317, 95% CI: 1.158–1.497), those with BMI < 24 kg/m² (IRR = 1.646, 95% CI: 1.335–2.032), non-smokers (IRR = 1.536, 95% CI: 1.285–1.836), and non-drinkers (IRR = 1.587, 95% CI: 1.276–1.973). Similarly, in the two-exposure models including heat exposure and H₂S, the positive association between heat exposure and high TC also appeared more pronounced among females (IRR = 2.347, 95% CI: 1.736–3.173), individuals aged ≥ 35 years (IRR = 1.385, 95% CI: 1.215–1.578), those with BMI < 24 kg/m² (IRR = 1.607, 95% CI: 1.288–2.004), non-smokers (IRR = 1.602, 95% CI: 1.344–1.911), and non-drinkers (IRR = 1.647, 95% CI: 1.323–2.051).

Fig. 2.

Subgroup-specific associations of heat exposure with high TC across demographic and occupational strata in two-exposure models. Bold indicates statistical significance (P < 0.05)

Sensitivity analyses for healthy worker bias

Table 4 presents the results of sensitivity analyses examining the potential impact of healthy worker bias. When stratified by employment duration, the inverse association between heat exposure and dyslipidemia was more pronounced among workers with ≥ 5 years of employment (IRR = 0.82, 95% CI: 0.75–0.89 for TG) compared to those with < 5 years (IRR = 0.91, 95% CI: 0.84–0.98), consistent with a cumulative healthy worker survivor effect. In analyses restricted to participants with normal lipid levels at baseline, the inverse associations were attenuated but remained statistically significant for TG (IRR = 0.88, 95% CI: 0.82–0.94) and LDL-C (IRR = 0.91, 95% CI: 0.85–0.97). Exclusion of workers with < 2 years of follow-up did not materially change the effect estimates (e.g., for TG: IRR = 0.84, 95% CI: 0.78–0.90).

Table 4.

Sensitivity analyses for healthy worker bias: associations between heat exposure and dyslipidemia

Analysis	TC	TG	LDL-C	HDL-C
Stratified by employment duration
< 5 years (n = 2,845)	0.95 (0.88–1.03)	0.91 (0.84–0.98)	0.94 (0.87–1.02)	0.97 (0.94–1.00)
≥ 5 years (n = 4,066)	0.89 (0.82–0.96)	0.82 (0.75–0.89)	0.88 (0.81–0.95)	0.95 (0.91–0.99)
Restricted to baseline normal	0.92 (0.86–0.98)	0.88 (0.82–0.94)	0.91 (0.85–0.97)	0.98 (0.95–1.01)
Excluding follow-up < 2 years	0.94 (0.88–1.00)	0.84 (0.78–0.90)	0.93 (0.87–0.99)	0.96 (0.93–1.00)

Data are presented as IRR (95% CI). All models adjusted for age, gender, BMI, smoking, and alcohol consumption. Bold indicates statistical significance (P < 0.05)

Lagged exposure analysis for healthy worker effect

To further assess whether the inverse associations observed in the primary models might be influenced by contemporaneous job assignment or early-career selection processes, we conducted lagged exposure analyses using prior-year exposure status to predict lipid abnormalities at the subsequent annual examination among workers with consecutive annual follow-up. As shown in Table 5, the inverse association between heat exposure and high TC observed in the primary concurrent models was attenuated and no longer statistically significant in the lagged analysis (IRR = 0.945, 95% CI: 0.817–1.092). In contrast, the inverse association with high TG remained statistically significant (IRR = 0.883, 95% CI: 0.818–0.954), whereas the association with high LDL-C was borderline positive but did not reach statistical significance (IRR = 1.101, 95% CI: 0.996–1.217). No statistically significant association was observed for low HDL-C (IRR = 0.573, 95% CI: 0.302–1.088).

Table 5.

Lagged associations of prior-year occupational heat exposure with dyslipidemia in consecutive annual observations

Outcome	IRR (95% CI) for prior-year heat exposure	P value
High TC	0.945 (0.817–1.092)	0.444
High TG	0.883 (0.818–0.954)	0.002
High LDL-C	1.101 (0.996–1.217)	0.06
Low HDL-C	0.573 (0.302–1.088)	0.089

Results were obtained from negative binomial GEE models. Prior-year heat exposure was used to predict lipid abnormalities at the subsequent examination. Models were adjusted for prior-year age, prior-year BMI, prior-year smoking status, prior-year drinking status, and examination year

We further evaluated prior-year heat exposure in two-exposure lagged models for high TC (Table 6). After simultaneous adjustment for prior-year gasoline exposure, prior-year heat exposure was no longer associated with high TC (IRR = 0.948, 95% CI: 0.819–1.097), and prior-year gasoline exposure was also not significantly associated with high TC (IRR = 0.859, 95% CI: 0.696–1.060). Similarly, after simultaneous adjustment for prior-year H₂S exposure, prior-year heat exposure remained non-significant (IRR = 0.954, 95% CI: 0.824–1.105), whereas prior-year H₂S exposure showed an inverse association with high TC (IRR = 0.814, 95% CI: 0.682–0.972). Overall, these lagged analyses did not support the positive high-TC associations observed in the concurrent two-exposure models and suggest that the primary inverse or paradoxical associations may have been influenced, at least in part, by healthy worker selection processes or other time-related sources of bias.

Table 6.

Lagged two-exposure models for high total cholesterol in consecutive annual observations

Model	Variable	IRR (95% CI)	P value
Prior-year heat + prior-year gasoline	Prior-year heat	0.948 (0.819–1.097)	0.474
	Prior-year gasoline	0.859 (0.696–1.060)	0.157
Prior-year heat + prior-year H₂S	Prior-year heat	0.954 (0.824–1.105)	0.533
	Prior-year H₂S	0.814 (0.682–0.972)	0.023

Exposure status at the previous annual examination was used to predict high TC at the subsequent examination. Models were adjusted for prior-year age, prior-year BMI, prior-year smoking status, prior-year drinking status, and examination year

Discussion

This study examined the associations between occupational hazard exposure and dyslipidemia among petrochemical workers, a population concurrently exposed to physical heat and chemical hazards such as gasoline and hydrogen sulfide. Dyslipidemia is a common occupational health concern in this workforce, with a reported prevalence of 41.4% in India and 45.4% in Iran [16, 36]. Occupational heat exposure is closely linked to increased cardiovascular risk [8], and combined exposure to heat and chemical pollutants may exacerbate lipid metabolism disorders [37]. However, evidence remains limited on whether the association between occupational heat exposure and dyslipidemia differs according to co-exposure to petrochemical hazards. In our study, a stronger positive association with high TC was observed for occupational heat exposure in two-exposure models that additionally included gasoline or H₂S exposure. These findings suggest that the association between heat exposure and high TC may differ according to selected chemical co-exposures. However, these results should be interpreted cautiously and should not be taken as conclusive evidence of biological synergy.

Stronger positive associations were observed in several strata, particularly among women, workers aged ≥ 35 years, non-smokers, and non-drinkers. However, these subgroup findings should be interpreted cautiously and do not by themselves establish true susceptibility. This study fills a critical gap by providing real-world evidence for co-exposure-related association patterns. Further research on underlying mechanisms is needed to inform targeted prevention strategies for dyslipidemia and related cardiovascular diseases (CVDs) in occupational populations with complex exposure profiles.

Relationship between occupational heat exposure and dyslipidemia

CVDs are a major health burden in the petrochemical industry [38]. Dyslipidemia, a well-established risk factor for CVDs, plays a pivotal role in its onset and progression [39, 40]. Regular occupational health surveillance is therefore important for early identification of lipid abnormalities in petrochemical workers. To our knowledge, this study is the first of its kind to evaluate the independent associations of occupational heat exposure with serum lipid abnormalities and to examine these associations in models that additionally accounted other occupational hazards among petrochemical workers.

Given the characteristics of petrochemical production, occupational heat exposure is one of the most prominent occupational hazards in this industry. Previous evidence suggests that heat exposure may increase cardiovascular risk partly through dysregulation of lipid metabolism [41]. For example, Vangelova et al. [15] examined 102 male industrial workers in Bulgarian ceramic foundries and found that heat exposure increased the risk of elevated total cholesterol (TC). Similarly, a study of 545 Bulgarian male workers reported a higher prevalence of dyslipidemia among middle-aged and older workers exposed to both heat and noise [19].

However, the association between occupational heat exposure and dyslipidemia remains inconsistent across studies. In our single-exposure analyses, heat exposure alone was inversely associated with HDL-C, LDL-C, and TG levels, which aligns with existing experimental and observational evidence. Yamamoto et al. reported that in healthy young men, HDL-C increased under moderate heat, while TC, TG, and LDL-C decreased under severe heat stress [42]. Similarly, in a prospective cohort of Chinese mining workers, Shan et al. observed that each 5 °C rise in mean temperature was associated with reduced TC, TG, and LDL-C [43]. Collectively, these results indicate that heat may influence lipid profiles through complex thermoregulatory and metabolic pathways, which warrants further mechanistic exploration.

Healthy worker effect: a potential explanation for the inverse association

Despite inconsistent findings on the heat-dyslipidemia association, the inverse relationship observed in our study appears paradoxical and may be partly explained by the healthy worker effect. The healthy worker effect is a well-recognized bias in occupational epidemiology, characterized by healthier individuals being more likely to enter and retain employment in hazardous occupations [44]. It encompasses two related phenomena: healthy hire bias and healthy worker survivor bias, resulting from ongoing selection processes and time-varying confounding related to prior exposure [45]. Workers less susceptible to adverse health outcomes tend to accumulate greater exposure over time, which may distort exposure-outcome associations.

Several features of petrochemical work may amplify this bias. First, workers assigned to high-temperature operations are often subject to stricter fitness requirements. Second, heat-exposed workers may also be younger and healthier at baseline. In addition, annual occupational health surveillance may further reinforce this selection process. Workers with abnormal findings may be reassigned or receive medical intervention, thereby increasing the likelihood that healthier workers remain in heat-exposed jobs for longer periods.

To further evaluate the possibility of healthy worker bias, we conducted several sensitivity analyses. The stronger inverse associations observed among workers with longer employment duration support the presence of healthy worker survivor bias, as healthier workers may selectively accumulate exposure over time [44, 45]. However, restriction to participants with normal lipid levels at baseline attenuated some inverse associations, suggesting that baseline health status and selection processes may also contribute to the observed findings [46]. In contrast, the stability of effect estimates after excluding workers with short follow-up argues against substantial bias from early health-related attrition [47]. Collectively, these sensitivity analyses suggest that healthy worker bias may have contributed to the observed inverse associations, although it is unlikely to fully explain them. Thus, the inverse association between heat exposure and dyslipidemia in this study is likely, at least in part, a manifestation of the healthy worker effect. Future studies with quantitative exposure assessment and active follow-up of workers who leave employment are needed to fully disentangle selection effects from causal effects.

This pattern also suggests a cumulative selection process rather than a cumulative toxic effect of heat exposure itself. In occupational settings with sustained medical surveillance and job retention requirements, workers who remain in heat-exposed positions for longer periods are more likely to represent a progressively selected subgroup with greater underlying health resilience [44, 45]. Therefore, the stronger inverse associations observed with longer employment duration are more plausibly interpreted as evidence of cumulative healthy worker survivor bias. They should not be interpreted as evidence that prolonged heat exposure exerts a progressively protective effect on lipid metabolism.

Interaction between occupational heat exposure and selected chemical co-exposures

Notably, when heat exposure, one chemical hazard, and their cross-product term were included in the same model, patterns consistent with positive interaction on the multiplicative scale were observed for high TC in the models involving gasoline and H₂S. No similarly consistent pattern was observed for the other chemical hazards or lipid markers. Given that statistically significant associations were identified for only two of seven tested chemicals and for one of four lipid markers, these findings warrant cautious interpretation and may partly reflect chance findings due to multiple testing [48]. Accordingly, these results are more appropriately interpreted as evidence of possible multiplicative interaction or effect modification, rather than definitive evidence of synergistic effects.

An important interpretive issue is the apparent difference between the inverse associations observed for heat exposure in the single-exposure models and the positive heat coefficients observed in some two-exposure models. This pattern should not be interpreted as indicating that healthy worker selection is present in the single-exposure models but absent in the two-exposure models. Rather, selection processes are likely to operate across all models, and the observed differences in the direction and magnitude of the heat coefficient more plausibly reflect the combined influence of selection, co-exposure structure, and model specification. In petrochemical workplaces, heat exposure commonly co-occurs with chemical hazards such as gasoline and H₂S; thus, omission of these co-exposures from single-exposure models may affect the estimated heat coefficient, whereas their inclusion may alter both the magnitude and direction of the association. Accordingly, the positive association between heat exposure and high TC observed in some two-exposure models should be interpreted cautiously as exploratory and hypothesis-generating, rather than as conclusive evidence of biological interaction.

Several non-mutually exclusive explanations may account for the selective pattern observed for TC. First, different chemicals may have distinct biological targets and metabolic pathways. Cholesterol and triglyceride metabolism are regulated by distinct, albeit interconnected, pathways [49]. Hepatic cholesterol homeostasis is primarily controlled by the SREBP-2 (sterol regulatory element-binding protein 2) and LXR (liver X receptor) pathways. These pathways regulate cholesterol synthesis, uptake, and efflux [49, 50]. In contrast, triglyceride metabolism is more directly governed by SREBP-1 and PPAR (peroxisome proliferator-activated receptor) signaling. These pathways control fatty acid synthesis and oxidation [51, 52]. Both heat stress and certain chemical exposures, including volatile organic compounds in gasoline and H₂S, can induce oxidative stress and inflammatory responses. These processes may preferentially disturb cholesterol regulatory pathways, thereby producing a more detectable signal for TC than for other lipid markers [49].

Second, the null or inverse associations observed for TG, LDL-C, and HDL-C may partly reflect greater biological and behavioral variability in these markers, which can make adverse associations more difficult to detect in an occupational cohort. Compared with TG and HDL-C, TC is relatively more stable over time and less sensitive to short-term behavioral influences such as recent diet, physical activity, and alcohol intake [53, 54]. In the presence of healthy worker selection and residual confounding, this difference in temporal stability may contribute to the more consistent pattern observed for TC.

Third, heat stress and chemical exposures may exert heterogeneous effects across different lipid fractions, resulting in an observable association for TC but not for TG, LDL-C, or HDL-C. Heat stress has been reported to induce inflammatory responses and adaptive metabolic changes, including alterations in fatty acid metabolism and lipid homeostasis [55]. Experimental studies of co-exposure have also suggested that combined physical and chemical stressors can disrupt lipid regulation. This disruption may occur through activation of the hypothalamic-pituitary-adrenal (HPA) axis and suppression of cholesterol clearance pathways, including CYP7A1 and ABCG5/G8 [37]. Such effects may preferentially affect overall cholesterol balance while producing less consistent changes in other lipid fractions.

Fourth, TC may represent an earlier or more sensitive marker of metabolic disturbance in this occupational population. In the natural history of lipid dysregulation, isolated hypercholesterolemia may precede the development of broader combined dyslipidemia involving multiple lipid fractions [56]. Because this workforce is relatively young and subject to occupational selection, some workers may be at an early stage of metabolic disruption in which TC abnormalities are detectable before changes in TG, LDL-C, or HDL-C become evident. Longer follow-up may therefore be required to determine whether comparable associations emerge for other lipid parameters.

However, the absence of consistent signals across multiple comparisons increases the probability of Type I error, and these findings should be considered hypothesis-generating rather than conclusive [57]. Further studies with larger sample sizes, explicit correction for multiple testing, and formal interaction analyses are needed to determine whether the observed associations reflect a reproducible co-exposure-related effect or statistical artifact.

Heat stress disrupts lipid homeostasis through multiple biological pathways, including sympathetic activation, mitochondrial dysfunction, oxidative and inflammatory stress, and altered balance between lipid synthesis and oxidation [58, 59]. Gasoline is a complex mixture of volatile organic compounds (VOCs) such as benzene, toluene, ethylbenzene, and xylene (BTEX). These compounds are metabolized in the liver primarily by cytochrome P450 enzymes (CYP2E1 and CYP1A1) [60, 61]. This metabolism generates reactive oxygen species and electrophilic intermediates. These processes lead to oxidative stress, lipid peroxidation, impaired mitochondrial β-oxidation, and activation of lipogenic signaling pathways [62–64]. Against a background of heat stress, these overlapping disturbances may provide a plausible explanation for the positive association observed after adjustment for gasoline co-exposure.

The biological effects of H₂S are complex and depend on dosage, duration, and exposure context. At physiological levels, endogenous H₂S functions as a gaseous signaling molecule with antioxidant and lipid-regulatory properties [65]. However, chronic occupational exposure to exogenous H₂S may inhibit mitochondrial cytochrome c oxidase. This inhibition impairs ATP production and promotes reactive oxygen species accumulation [66, 67]. This process may shift the intracellular environment toward oxidative and inflammatory stress. Because heat exposure may also impair mitochondrial function and reduce fatty acid oxidation [68], concurrent exposure may intensify disturbances in hepatic lipid synthesis and clearance. Consistent with this broader concept, Jiang et al. reported that co-exposure to ozone and heat disrupted lipid homeostasis, accompanied by elevated stress hormone levels and activation of the HPA and sympathetic-adrenal-medullary axes [37].

Taken together, these findings suggest that the positive association observed for heat exposure with high TC in the models additionally accounting for gasoline or H₂S may reflect disturbances in cholesterol homeostasis rather than a uniform effect on overall lipid metabolism. They also raise the possibility that TC may be a relatively early and sensitive indicator of metabolic disturbance in this occupational setting. However, this interpretation remains exploratory, given the potential influence of healthy worker selection, residual confounding, dichotomous exposure classification, and multiple comparisons. Future studies incorporating quantitative exposure assessment, longer follow-up, lipoprotein subfraction measurements, and formal interaction testing will be necessary to clarify whether these associations are reproducible and to define the underlying mechanisms more rigorously.

Stratified analysis of subgroup-specific associations

In stratified analysis, the positive association between heat exposure and high TC in the two-exposure models appeared more pronounced among women, workers aged ≥ 35 years, non-smokers, and non-drinkers. However, these subgroup-specific associations should not be interpreted as evidence of intrinsic susceptibility. Rather, they are more likely to reflect differences in job assignments, workplace environments, and organizational constraints across subgroups. In petrochemical workplaces, task allocation and production-zone assignment often determine exposure patterns, smoking restrictions, and alcohol consumption policies [69–71]. Therefore, observed differences across gender and lifestyle strata may primarily reflect occupational context rather than biological differences.

Several lines of evidence suggest that occupational contextual factors may have contributed to the subgroup-specific patterns observed in this study. First, previous studies have shown that men and women with the same job title often perform different tasks, and may therefore experience different exposure profiles [69, 70]. For example, Messing et al. [69] found that job title alone may inadequately capture exposure status because task allocation often differs by gender. Similarly, Eng et al. [70] reported that, even within the same occupation, men were more likely to be exposed to physical and chemical hazards, whereas women faced more ergonomic stressors. A more recent review further confirmed that systematic gender-related differences in occupational exposures remain common across work settings [71].

Second, the subgroup in which the association appeared more pronounced was predominantly engaged in oil refining, catalytic reforming, or laboratory testing positions [72, 73]. These roles strictly prohibit smoking and alcohol consumption in production areas, explaining their non-smoking and non-drinking status. However, whether these positions entail systematically different chemical exposures than male-held jobs could not be determined from our dichotomous data and warrants further investigation.

Third, residual confounding by unmeasured factors, such as lifestyle behaviors, socioeconomic status, dietary patterns, physical activity, body composition, genetic predisposition, reproductive status, menopausal status, or cumulative exposure duration, may partially explain the observed associations [74]. Although we adjusted for available covariates, uncontrolled confounding cannot be excluded.

Notwithstanding these occupational explanations, biological mechanisms may also contribute, although they cannot be distinguished from occupational or residual confounding effects in this observational setting. Li et al. [75] reported that women had a higher risk of dyslipidemia than men in a nationwide Chinese sample. Fernandez et al. [76] suggested that increased dyslipidemia risk in middle-aged and older women is associated with postmenopausal estrogen decline. However, the age category of ≥ 35 years used in our subgroup analysis is too broad to distinguish occupational exposure-related effects from natural physiological transitions across the reproductive and perimenopausal period. Estrogen enhances hepatic uptake of TC and LDL-C, promotes HDL-C synthesis, facilitates bile acid secretion, and accelerates cholesterol clearance; reduced estrogen levels or menopause may increase dyslipidemia risk [77]. Therefore, both occupational and biological explanations remain plausible, but neither can be isolated with the available data.

Regarding smoking, prior evidence shows that smokers tend to have more adverse lipid profiles than non-smokers [78]. This contrasts with our finding that non-smoking female workers exhibited higher TC abnormality risk. In this occupational setting, non-smoking and non-drinking status are more plausibly interpreted as proxies for assignment to specific production-zone jobs where such behaviors are restricted and where exposure patterns (heat, chemical co-exposure, and PPE use) may differ. Therefore, the observed association is more likely to reflect occupational context rather than individual lifestyle factors or intrinsic susceptibility. Residual confounding cannot be excluded, and the absence of detailed task-level and PPE information limits our ability to distinguish behavioral from occupational explanations.

Strengths and limitations

This study has several strengths, including a retrospective cohort design, 9 years of occupational health examination data, a relatively large sample size, and the use of GEE models to evaluate the associations of occupational heat exposure among petrochemical workers. The repeated-measure design also allowed us to examine these associations in both primary and lagged analyses and to further explore patterns observed in models that additionally accounted for selected co-occurring workplace hazards.

Nevertheless, several limitations should be considered. (1) All occupational exposure variables in this study were dichotomized (exposed vs. non-exposed) based on occupational health records. Although occupational heat exposure was defined with reference to the WBGT threshold specified in occupational health standards, the actual exposure classification in this study relied on administrative hazard records rather than repeated individual-level quantitative measurements. As a result, this approach may not fully capture variability in specific job tasks, time spent in exposed environments, or the use of PPE, and may therefore introduce exposure misclassification. In addition, the dichotomous exposure definition precluded evaluation of exposure-response relationships, as we were unable to quantify the intensity (e.g., actual heat stress level measured by WBGT) or duration (e.g., cumulative exposure years or hours per shift) of individual exposures. Consequently, our analysis may obscure heterogeneity in exposure levels and cannot distinguish differences within exposure categories defined as exposed vs. non-exposed based on occupational health records. This limitation could potentially attenuate true associations or reduce our ability to detect effects occurring only at higher exposure levels, and also limited the clinical interpretability of the estimated IRRs. Future studies should incorporate quantitative exposure assessment methods, such as personal monitoring devices, task-based exposure assessment, detailed exposure matrices, and information on PPE use, to enable dose-response analysis and more precisely characterize the health risks associated with occupational heat and chemical exposures. In addition, smoking and drinking status may have partly reflected worksite restrictions and production-zone assignment rather than purely personal lifestyle behaviors; (2) Only process-generated workplace heat exposure was considered, with no inclusion of ambient meteorological condition changes that may affect blood lipid profiles; (3) No data on dietary habits, lipid-lowering medication use or other metabolic covariates were collected, limiting refined adjustment and assessment of the heat-dyslipidemia association; (4) The use of negative binomial GEE models for binary outcomes, while methodologically justified for estimating incidence rate ratios and addressing overdispersion, is less conventional and may complicate interpretation. However, prior methodological studies suggest that, under comparable settings, the estimated incidence rate ratios are generally similar to those obtained from log-binomial or Poisson GEE models with robust standard errors. Future studies could further assess the robustness of these findings through sensitivity analyses using alternative modeling approaches [79, 80]; (5) We conducted a series of two-exposure models in which heat exposure was evaluated together with one chemical hazard at a time across four lipid markers, yielding 28 exposure-outcome comparisons in total without correction for Type I error. Given the number of statistical comparisons performed, the possibility that some observed associations reflect chance findings cannot be excluded; and (6) The interpretation of the two-exposure models also requires caution. In these models, heat exposure, one chemical hazard, and their cross-product term were included simultaneously. The analyses assessed interaction on the multiplicative scale rather than additive interaction, and the results should not be interpreted as definitive evidence of biological synergy. Therefore, the positive association between heat exposure and high TC in some two-exposure models should be regarded as exploratory and hypothesis-generating.

Conclusions

The inverse association observed between occupational heat exposure and dyslipidemia in the single-exposure models may be partly attributable to the healthy worker effect. In two-exposure models that additionally accounted for gasoline or H₂S, heat exposure was positively associated with elevated total cholesterol. However, this finding should be interpreted cautiously as exploratory evidence of possible interaction on the multiplicative scale, rather than as definitive evidence of a true joint effect. Further studies with more refined exposure assessment, longer follow-up, and validation in other occupational populations are needed to clarify and confirm these associations.

Abbreviations

TC

Total cholesterol

LDL-C

Low-density lipoprotein cholesterol

HDL-C

High-density lipoprotein cholesterol

TG

Triglyceride

IRR

Incidence Rate Ratio

LHDL-C

Hypo-high-density lipoproteinemia

HLDL-C

Hyper-low-density lipoproteinemia

GEE

Generalized estimating equations

BMI

Body Mass Index

Authors’ contributions

Yilin Zhang and Yifeng Chen: Methodology, Data analysis, Writing – original draft, Data curation and management. Xiaoyun Li： Data curation and data collection, Data analysis. Qingyu Li: Writing – review & editing, Methodology, Data analysis. Yan Yang： Writing – review & editing, Methodology, Data curation. Shanshan Du, Fei He, Zihu Lv： Supervision, Conceptualization, data collection and management. Weimin Ye, Wei Zheng, Jianjun Xiang: Conceptualization, Methodology, Writing – review & editing, Supervision, Formal analysis, Funding acquisition.

Funding

This study was supported by the Minjiang Scholar Start-up Research Fund of Fujian Province (Grant No. 2019-9202001001) and 2021 Natural Science Foundation of Fujian Province of China (2021J01722).

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Ethics Committee of Fujian Medical University (approval No. 2022 − 111). The research was conducted in accordance with the ethical guidelines and principles outlined in the Declaration of Helsinki. The study was conducted in accordance with the ethical guidelines and principles outlined in the Declaration of Helsinki. This study was based on retrospectively collected occupational health surveillance data that had been de-identified before delivery. The Ethics Committee waived the requirement for individual informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.