Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Mar 2.
Published in final edited form as: Ecotoxicol Environ Saf. 2025 Feb 14;291:117897. doi: 10.1016/j.ecoenv.2025.117897

Potential causal associations of long-term exposure to PM2.5 constituents and all-cause mortality: Evidence from the Pearl River Cohort study

Huanle Cai a,1, Yuqin Zhang a,1, Jing Wei b, Wayne R Lawrence c, Shirui Chen a, Huanhuan Cheng d, Hui Tang a, Dan Chen a, Ying Wang a, Jing Gu a,*, Wangjian Zhang a,*, Yuantao Hao e,f,**
PMCID: PMC11871989  NIHMSID: NIHMS2058416  PMID: 39954626

Abstract

Evidence of the potential causal effect of PM2.5 and its constituents on all-cause mortality based on large population cohort is still limited. Based on a large scale cohort of 341,098 participants in southern China, we developed a marginal structure Cox model based on inverse probability weighting, an established causal inference approach, to evaluate the potential causal associations between PM2.5 constituents and all-cause mortality, taking into account time-varying covariates. Additionally, we further explored the modifying effects of demographic and lifestyle characteristics on these associations. For each IQR increase in black carbon and organic matter, the risk of all-cause mortality increased by 51 % (95 % CI: 40–62 %) and 52 % (95 % CI: 40–58 %), followed by ammonium, nitrate and sulfate (HR = 1.32, 1.38 and 1.39, respectively). Individuals < 65 years, unmarried, urban medical insured, not consuming alcohol, or not exercising were potentially more susceptible to the adverse effects for most PM2.5 constituents (P for interaction < 0.10). The HR estimates for these subgroups ranged from 1.51 to 2.10 for black carbon, 1.40–1.71 for nitrate, 1.42–1.53 for sulfate, and 1.52–1.77 for organic matter. Various sensitivity analyses verified the robustness of our findings. This study presented compelling evidence of potential causal links between PM2.5 constituents and all-cause mortality.

Keywords: Causal inference, PM2.5 constituents, Large cohort, All-cause mortality

1. Introduction

Epidemiological research have demonstrated that atmospheric fine particulate matters (PM) exposure is related to a greater risk of diverse detrimental health results, such as cardiovascular diseases, respiratory diseases, tumors, and mutations (Mukherjee and Agrawal, 2018; Chen and Hoek, 2020; Orellano et al., 2020). Based on the Global Burden of Diseases (GBD) study (2019), over 4 million deaths globally could be resulted from PM exposure in that year, where 1.4 million of these deaths occurred in China (Collaborators, 2020). Potential mechanisms underlying the negative health effects of particulate matter exposure include intracellular oxidative stress, mutation and genotoxic reactions, and systemic inflammatory responses (Feng et al., 2016).

The risk of mortality attributed to PM2.5 exposure was highly heterogeneous across regions, which might be due to the variation in its constituents (Liu et al., 2019). PM2.5 is a composite mixture consisting of several chemical constituents, like black carbon, ammonium, nitrate, sulfate, and organic matter (Kang et al., 2023). Black carbon and organic matter primarily result from incomplete fuel combustion, which is more prevalent in areas with heavy traffic and industrial activities (Pateraki et al., 2020). Ammonium, nitrate, and sulfate are secondary pollutants formed through chemical reactions in the air involving sulfur dioxide and nitrogen oxides, originating from biomass and diesel combustion, and automobile exhaust emissions (Pateraki et al., 2020). The health effects of a specific constituent of PM2.5 are diverse, while the overall toxicity of PM2.5 largely depends on its composition (Wang et al., 2022; Shi et al., 2023; Liu et al., 2021). A prior study based on 16 cities in China showed that the toxicity of PM in the Pearl River Delta region was higher than that in other regions, which may be related to the heavy traffic emissions, weather patterns, and susceptibility of residents in the Pearl River Delta region (Chen et al., 2012). Although several studies have investigated the relationship between PM2.5 constituents and mortality in other regions of China (Chen et al., 2021, 2018; Wang et al., 2019; Fu et al., 2023), there is still a dearth of solid evidence from the Pearl River Delta area, particularly based on large population cohorts.

Moreover, a large proportion of existing studies generally are supported by traditional methods of association assessment, the estimates from which are typically “conditional effects” obtained from regression models with covariates being directly incorporated (Dominici and Zigler, 2017; Janes et al., 2010). Given the variation in confounders across the individuals, the “conditional effect” often tends to be closer to the mean effect at the individual level (Wang et al., 2023). However, compared to these individual-level estimates, the population-level effect estimates generally have more policy implications, thus, are more preferred. Recent advances in causal inference approaches for the observational data enable the evaluation of the population-level effect estimates. By mimicking the randomized controlled trails, a pseudo-population would be created with confounders being homogenized across exposure groups. Effect estimates generated by comparing across the pseudo groups are considered having potential causal implications. Nevertheless, such evidence remains restricted in current studies.

This study adopted a causation inference method to explore the potential causal link between long-term PM2.5 constituents exposure and all-cause mortality based on the Pearl River Cohort. We further assessed the potential modifying effects of demographic features and living habits.

2. Methods

2.1. Study population and outcome

The study utilized a large population cohort from the Major Science and Technology Projects of the 11th and 12th Five-Year Plans in China, which oriented to establish national demonstration areas for community-level collaborative innovation projects (Ruan et al., 2019; Wang et al., 2024a). And the participants in this cohort were selected on a community scale in southern China by stratified random cluster sampling. All enrolled subjects were required to undertake a comprehensive survey to obtain research information. Due to the accessibility of PM2.5 constituents exposure concentrations, the study focused on a sub-cohort enrolled after January 1, 2013, consisting of 341,098 participants from Guangzhou and Zhongshan areas. Ethical approval was obtained by us from the Institutional Review Board at Sun Yat-sen University (L2017–030). All the individual participants in the study provided their informed consent.

The mortality data were obtained from the Centers for Disease Control and Prevention Death Registry in Guangzhou and Zhongshan. The length of the follow-up period was set as the time between registration and occurrence of death or the termination of the study on December 31, 2020.

2.2. Exposure and covariates

The main exposure in this study was PM2.5 concentration and its constituents, namely black carbon, ammonium, nitrate, sulfate, and organic matter. The annual average concentrations of these pollutants, having a spatial resolution of 1 km × 1 km, were obtained from the ChinaHighAirPollutants (CHAP) datasets. The data collection and modeling process have been previously outlined in detail (Wei et al., 2023, 2021, 2020). The model estimation of these constituents’ concentrations are well validated and widely used. In general, in relation to 126 ground monitoring stations distributed nationwide, the cross-validation R2 values of each constituent ranged from 0.71 to 0.75 and the root-mean-square error (RMSE) was within the scope of 4.3–6.6 (Wei et al., 2023). Average annual exposure concentration for each pollutant was allocated to each participant in accordance with their residential address.

We utilized a directed acyclic graph (DAG) to detect confounding factors (Lipsky and Greenland, 2022). Mapping the DAG was informed by an extensive literature review and our prior research experience, as depicted in Fig. S2. Based on the insights from the DAG, the factors included sociodemographic factors (age, gender, education, marital status, ethnicity, and medical insurance) and lifestyle variables (smoking, alcohol consumption, and exercise). Information on participants’ sociodemographic and lifestyle factors was gathered through direct-contact interviews with the assistance of well-trained nurses using computer-based questionnaires. These variables included age group (< 65, ≥ 65), gender (male, female), education level (elementary, middle school, high school, and college), marital status (single, married, and widowed/divorced), ethnicity (Han or other), medical insurance (urban workers, non-working urban residents, new rural cooperative, or other), smoking habits (No or Yes), with Yes signifying at least one cigarette daily for a minimum duration of 6 months, alcohol consumption (No or Yes), with No signifying either no alcohol consumption or only a limited amount consumed within the past year, and exercise habits (No or Yes).

2.3. Statistical analysis

In this study we employed a marginal structure Cox model (MSM) utilizing inverse probability weighting (IPW) for assessing the potential causal relationship between each IQR increase in concentrations of PM2.5 and its constituents and all-cause mortality. To account for the decreasing trend in the concentration of PM2.5 and its constituents over time, we treated them as time-dependent variables (Zhang et al., 2018). The IPW approach utilized the concept of counterfactuals to simulate observational data in the form of a randomized controlled trial in order to create pseudo-populations and achieve the effect of balancing the observed confounding factors (Wu et al., 2020a; Cole and Hernan, 2008). The stabilized weights for the exposures were calculated as the ratio of the marginal probability of exposure to the generalized propensity score (GPS) (Robins et al., 2000). The GPS was obtained through regression of exposure variables against potential confounding factors. In order to address extreme weights, we capped the weights at 10 for those with a value exceeding 10 (Gruber et al., 2022). Moreover, we employed three different regression methods for calculating GPS: linear model (LM), generalized estimating equation (GEE), and gradient boosting machine learning (ML). The balance of covariates in the pseudo-population was assessed using the mean absolute correlation (AC), where AC values under 0.1 indicated the high quality of the simulated randomized controlled trial (Fig. S1) (Austin, 2019). Based on the results, we selected the GEE-IPWs method, which attained the best balance in covariates, as our main analysis model.

We adopted a sequential modeling strategy as follows:

Model 0: the crude time-dependent Cox model including either PM2.5 or its constituents as the sole variable.

Model 1: model 0 with additional adjustment for age and gender.

Model 2: model 1 with further adjustment for education level, marital status, ethnicity, medical insurance, smoking habits, alcohol consumption and exercise habits.

Model 3: model 2 weighted by GPS using GEE.

We evaluated collinearity in the final model using the generalized variance inflation factor (GVIF), ensuring that GVIF values were below 5. Building upon Model 3, we carried out a stratified analysis according to sociodemographic and lifestyle variables to assess potential modification effects. Notably, upon applying the weights of Model 3 in the total population to the stratified analyses, the balance of the covariates within subgroups was also achieved, as indicated by the AC values for the exposure against confounders in each subgroup (Table S1, all AC < 0.1). For imputing missing data, we employed the R multivariate imputation by chained equation (MICE) package (Zhang, 2016).

In order to guarantee the robustness of our results, several sensitivity analyses were performed. First, the final model was applied to the complete datasets without imputing missing values to assess the impact of data imputation. The characteristics of participants in the complete datasets are shown in Table S2. Second, we utilized IPW obtained through all three weighting methods in the final model. Third, a multi-constituent model analysis was conducted to evaluate the influence between constituents with low correlations (r < 0.7). In the multi-constituent model, all constituents with low correlations to the major constituent were adjusted as covariates based on the single-constituent model (Model 3) (Tao et al., 2023). Fourth, we also calculated the E-value to evaluate the implications of unmeasured confounding factors for the study results. A larger E-value indicated greater robustness of the results (Haneuse et al., 2019). The main outputs for these models were hazard ratio (HR) estimates for every IQR rise in PM2.5 and its constituents. All the analysis work was performed using R version 4.1.3. Two-sided P-values < 0.05 were deemed statistically significant. We also evaluated P-values < 0.10 as the threshold for statistical interaction (Selvin, 2004).

3. Results

During the 8-year follow-up period, among the 341,098 participants, a total of 10,683 deaths occurred, representing a mortality rate of 3.13 % (Table 1). Male participants accounted for 49.9 % of the total sample. The mean age of the all-cause deaths group was 70.54 ± 14.58 years, while the survived group had an average age of 42.69 ± 16.07 years. The average exposure concentration of PM2.5 among participants in 2013 was 52.14 ± 3.10 μg/m3, gradually decreasing each year to 22.25 ± 1.77 μg/m3 by 2020. The exposure concentration trend was similar for PM2.5 constituents (Table 2).

Table 1.

sociodemographic and lifestyle factors of all participants.

Overall
(N = 341,098)
N (%)		 Survived (n = 330,415) n (%) All-cause death
(n = 10,683)
n (%)		 P
Age (m±s) 43.57 ± 16.75 42.69 ± 16.07 70.54 ± 14.58 < 0.001
Age group < 0.001
 < 65 299,126 (87.7) 295,635 (89.5) 3491 (32.7)
≥ 65 41,972 (12.3) 34,780 (10.5) 7192 (67.3)
Gender < 0.001
 Male 170,243 (49.9) 164,011 (49.6) 6232 (58.3)
 Female 170,855 (50.1) 166,404 (50.4) 4451 (41.7)
Education < 0.001
 Elementary 39,326 (11.5) 35,959 (10.9) 3367 (31.5)
 Middle 99,047 (29.0) 95,549 (28.9) 3498 (32.7)
 High 153,413 (45.0) 150,210 (45.5) 3203 (30.0)
College 49,312 (14.5) 48,697 (14.7) 615 (5.8) < 0.001
Marital status
 Single 78,385 (23.0) 77,648 (23.5) 737 (6.9)
 Married 254,997 (74.8) 246,225 (74.5) 8772 (82.1)
 Widowed & Divorced 7716 (2.3) 6542 (2.0) 1174 (11.0)
Ethnic < 0.001
 Han 334,453 (98.1) 323,880 (98.0) 10,573 (99.0)
 Others 6645 (1.9) 6535 (2.0) 110 (1.0)
Medical insurance < 0.001
 Urban workers 117,033 (34.3) 111,269 (33.7) 5764 (54.0)
 Non-working urban residents 72,394 (21.2) 69,331 (21.0) 3063 (28.7)
 New rural cooperative 34,616 (10.1) 34,181 (10.3) 435 (4.1)
 Others 117,055 (34.3) 115,634 (35.0) 1421 (13.3)
Smoke < 0.001
 No 331,680 (97.2) 321,634 (97.3) 10,046 (94.0)
 Yes 9418 (2.8) 8781 (2.7) 637 (6.0)
Alcohol < 0.001
 No 320,280 (93.9) 310,652 (94.0) 9628 (90.1)
 Yes 20,818 (6.1) 19,763 (6.0) 1055 (9.9)
Exercise < 0.001
 No 98,193 (28.8) 93,642 (28.3) 4551 (42.6)
 Yes 242,905 (71.2) 236,773 (71.7) 6132 (57.4)
Open in a new tab

Table 2.

Annual exposure concentration of PM2.5 and its constituents (mean ± sd).

PM2.5 Black carbon Ammonium Nitrate Sulfate Organic matter
2013 52.14 ± 3.10 4.25 ± 0.36 5.82 ± 0.18 9.24 ± 0.65 11.2 ± 0.60 19.8 ± 1.93
2014 45.11 ± 4.21 3.76 ± 0.39 5.16 ± 0.31 8.02 ± 0.87 9.91 ± 1.00 16.1 ± 2.59
2015 38.18 ± 2.51 3.44 ± 0.38 4.51 ± 0.20 6.54 ± 0.52 8.80 ± 0.70 13.1 ± 1.67
2016 34.55 ± 2.19 3.21 ± 0.33 4.06 ± 0.20 5.81 ± 0.51 7.89 ± 0.63 11.7 ± 1.55
2017 35.40 ± 1.46 3.22 ± 0.27 4.13 ± 0.22 6.34 ± 0.46 8.00 ± 0.35 12.1 ± 1.02
2018 33.05 ± 2.65 3.12 ± 0.36 3.93 ± 0.18 6.10 ± 0.65 7.45 ± 0.53 10.8 ± 1.85
2019 29.57 ± 1.62 2.96 ± 0.29 3.49 ± 0.16 5.33 ± 0.42 7.12 ± 0.42 9.18 ± 1.26
2020 22.25 ± 1.77 2.41 ± 0.29 2.65 ± 0.14 4.10 ± 0.36 5.62 ± 0.49 6.23 ± 1.18
Open in a new tab

Fig. 1 displays the HR estimates for the relationship between PM2.5 constituents and all-cause mortality risk in both the marginal structure Cox model and the ordinary time-dependent Cox model. The HR estimate values in all Models indicated that PM2.5 and each constituent were risk factors for all-cause death. In the final model (Model 3), the HR for risk of all-cause mortality was 1.45 (95 % CI: 1.35–1.56) for each IQR rise in PM2.5 concentration. The HRs for black carbon, ammonium, nitrate, sulfate, and organic matter were 1.51 (95 % CI: 1.40–1.62), 1.32 (95 % CI: 1.26–1.38), 1.38 (95 % CI: 1.32–1.46), 1.39 (95 % CI: 1.31–1.45), and 1.52 (95 % CI: 1.40–1.58) for each IQR rise in the concentration, respectively. Comparatively, the ordinary time-dependent Cox model (Model 2) estimated the risk of all-cause mortality attributed to PM2.5 and its constituents to be a little lower than the estimates obtained from the main model (Model 3). Table 3 presents the estimates for the remaining models.

Fig. 1.

Fig. 1.

Open in a new tab

HR and 95 % CI of PM2.5 and its constituents to all-cause mortality in two different models. The red represents the marginal structure Cox model, which is weighted based on the ordinary Cox model using GPS obtained by GEE (Model 3). And the blue represents the time-dependent Cox model (Model 2).

Table 3.

Hazard ratio (HR) of PM2.5 and its constituents to all-cause death.

Model 0 Model 1 Model 2 Model 3
HR 95 % CI HR 95 % CI HR 95 % CI HR 95 % CI
PM2.5 2.72*** (2.54–2.90) 1.35*** (1.35–1.45) 1.35*** (1.25–1.35) 1.45*** (1.35–1.56)
Black carbon 3.53*** (3.38–3.69) 1.33*** (1.27–1.39) 1.33*** (1.26–1.41) 1.51*** (1.40–1.62)
Ammonium 1.13*** (1.07–1.18) 1.34*** (1.28–1.40) 1.30*** (1.25–1.36) 1.32*** (1.26–1.38)
Nitrate 1.97*** (1.88–2.08) 1.40*** (1.34–1.46) 1.31*** (1.25–1.38) 1.38*** (1.32–1.46)
Sulfate 2.55*** (2.47–2.66) 1.26*** (1.20–1.30) 1.23*** (1.17–1.28) 1.39*** (1.31–1.45)
Organic matter 3.48*** (3.37–3.60) 1.29*** (1.24–1.40) 1.29*** (1.19–1.34) 1.52*** (1.40–1.58)
Open in a new tab

Model 0: õne of PM2.5 or its constituents.

Model 1: model 0 +age+gender.

Model 2: model1 +education+marital status+ethnic+medical insurance+smoke+alcohol+exercise.

Model 3: model 2 weighted by generalized estimating equation (GEE).

***:

P < 0.001.

In the subgroup analysis (Fig. 2), we observed significant modification effects for age (all P for interaction < 0.001), marital status (except for sulfate, the other P for interaction < 0.10), medical insurance (all P for interaction < 0.05), alcohol consumption (all P for interaction < 0.10) and exercise (all P for interaction < 0.001). Apart from ammonium, the impact on the risk for all-cause mortality was greater in the < 65 years old group (HR of <65 vs. ≥65, black carbon: 1.92 vs. 1.53, nitrate: 1.48 vs. 1.34, sulfate: 1.51 vs. 1.46, organic matter: 1.71 vs. 1.58). The risk of all-cause mortality related to PM2.5 constituents was lowest among married individuals, with HR = 1.44 for black carbon, HR = 1.36 for nitrate, and HR = 1.46 for organic matter, followed by the HR estimates in the single group and the widowed & divorced group. Furthermore, participants with medical insurance for non-working urban residents showed the highest susceptibility to the effects of nitrates (HR = 1.50, 95 % CI, 1.34–1.64), sulfates (HR = 1.53, 95 % CI, 1.39–1.68), and organic matter (HR = 1.64, 95 % CI, 1.46–1.84), followed by the estimates in the urban workers medical insurance group and other groups. Residents with urban workers medical insurance were most vulnerable to black carbon (HR = 1.87, 95 % CI, 1.66–2.10), followed by the estimates in the medical insurance group for non-working urban residents (HR = 1.45, 95 % CI, 1.28–1.64), and in the others medical insurance group (HR = 1.34, 95 % CI, 1.11–1.60). In addition to the above factors, we did not find modification effect when the estimates were stratified by other demographic features, such as gender and education.

Fig. 2.

Fig. 2.

Open in a new tab

The modification effects on the causal association between PM2.5 constituents and all-cause mortality.

Similarly, when the results were stratified by lifestyle factors, residents who did not consume alcohol exhibited increased susceptibility to PM2.5 constituents exposure (HR of No vs. Yes, black carbon: 1.51 vs. 1.29, ammonium: 1.34 vs. 1.13, nitrates: 1.40 vs. 1.18, sulfates: 1.42 vs. 1.01, organic matter: 1.52 vs. 1.04). Additionally, exercise showed a statistically significant interaction effect with each constituent. Individuals who did not exercise were more susceptible to black carbon (HR of No vs. Yes, 2.10 vs.1.41), sulfates (HR of No vs. Yes, 1.48 vs.1.36) and organic matter (HR of No vs. Yes, 1.77vs.1.40), while individuals who exercise were more susceptible to ammonium (HR of No vs. Yes, 1.27 vs.1.33) and nitrates (HR of No vs. Yes, 1.36vs.1.38). Effect modification regarding smoking was not observed.

The results obtained from the imputed data sets and complete data sets demonstrated the robustness of the findings (Table S3). Similarly, the results from the final model weighted by GPS using different methods (LM, ML, and GEE) were also consistent and reliable (Table S4). In the multi-constituent model, the HR of each main constituent remained significant, suggesting the results were robust compared with the single-constituent model (Table S5). Moreover, the E-values were all greater than 1, indicating that unmeasured confounding factors were unlikely to substantially change our observed findings (Table S6).

4. Discussion

In this large cohort study employing causal inference methods, we discovered a significant potential causal relationship between long-term PM2.5 constituents exposure and an elevated risk of 32–52 % in all-cause mortality. Among PM2.5 constituents, black carbon and organic matter had the highest health risk, followed by ammonium, nitrate and sulfate. Furthermore, our findings indicate potential modification effects of age, marital status, medical insurance, alcohol consumption, and exercise.

We observed significant associations between long-term exposure to PM2.5 constituents and the increased mortality, which was consistent with previous findings (Hvidtfeldt et al., 2019; Goobie et al., 2022; He et al., 2024; Ostro et al., 2010). When the effect size of this study is converted to reflect a 1 μg/m3 increase rather than an IQR increase in the exposure, it ranges from 1.10 to 1.83, also consistent with the range of estimates in existing studies. For instance, Ostro et al. (2010) reported in a study in California that the HRs for all-cause mortality ranged from 1.06 to 1.93 following each 1 μg/m3 increase in long-term exposure to PM2.5 constituents. Similarly, a Danish study suggested HRs of mortality ranging from 1.07 to 4.66 per 1 μg/m3 increment in the concentration of PM2.5 constituents (Hvidtfeldt et al., 2019). Furthermore, several studies have reported on the health effects of long-term exposure to PM2.5 constituents from the perspective of cause-specific mortality, and the estimates in these studies tended to be even greater than the estimates in this study. For example, a global multi-center study reported that each 1 μg/m3 increase in the concentration of PM2.5 and its constituents including sulfate, nitrate, and ammonium, was associated the HRs of mortality of 1.18, 8.02, 3.78, and 50.99, respectively, among patients with fibrotic interstitial lung disease (Goobie et al., 2022). Another cohort study focusing on Chinese children found that among all the components, black carbon had the greatest impact on mortality in childhood-onset lupus nephritis, with an HR of 2.14 for each 1 μg/m3 increase in its concentration (He et al., 2024). These findings not only corroborated the connection between PM2.5 constituents and mortality but also underscored the global heterogeneity in the risk of mortality (Chen et al., 2024). Consequently, our study furnished compelling evidence regarding the impact of exposure to PM2.5 chemical constituents on all-cause mortality specifically within the Pearl River Delta region, contributing to the growing body of knowledge in this area.

We found that among the PM2.5 constituents, black carbon and organic matter had the highest association with the risk of all-cause mortality, which was consistent with previous findings (Wang et al., 2019; He et al., 2024; Yang et al., 2019) and was supported by plausible biological mechanisms. Black carbon and organic matter, which are major constituents of PM2.5, are directly released from the burning of fossil fuels and can adhere to ambient particulate matter (Liu et al., 2013). Black carbon has been linked to DNA methylation and oxidative stress (Niranjan and Thakur, 2017), which could contribute to cardiovascular damage and lung injury (Hvidtfeldt et al., 2019), thereby increasing the risk of suffering from cardiovascular diseases and the associated mortality (Yang et al., 2019). Similarly, organic matter has been associated with elevated blood pressure, raised heart rate variability, inflammation, and oxidative stress (Chen et al., 2021). These findings were consistent with epidemiological studies that suggested black carbon and organic matter increases risk of cardiovascular- and respiratory-related mortality (Yang et al., 2020). Ammonium, nitrate, and sulfate are the secondary pollutants formed through atmospheric reactions involving sulfur dioxide and nitrogen oxides (Pateraki et al., 2020). Nitrate could exert effect on the central nervous system by stimulating the hypothalamic-pituitary-adrenal (HPA) axis. This activation could either amplify or inhibit cardiometabolic responses. Consequently, nitrate demonstrated a more pronounced correlation with the onset of human cardiovascular diseases (Niu et al., 2018). Furthermore, research had indicated that nitrates were associated with lung inflammatory cells and respiratory tract damage (Zhang et al., 2021). Such associations could precipitate the onset and progression of respiratory diseases. Moreover, the acidity of nitrate and sulfate could directly induce airway hyper-reactivity and impair airway clearance, leading to lung injury and facilitating the uptake, conversion, and application of toxic substances (Yang et al., 2020). Notably, sulfate has been demonstrated to exhibit a positive correlation with biomarkers indicative of oxidative stress (Li et al., 2016). Cumulatively, these aspects constitute significant risk factors contributing to the development of cardiovascular disease (Yang et al., 2020). Ammonium, usually combined with nitrate and sulfate in the form of ammonium nitrate (NH4NO3) and ammonium sulfate((NH4)2SO4), had uncertain biological mechanisms (Hao et al., 2023).

When results were stratified by demographic characteristics, we observed effect modification by age, marital status, and medical insurance. The subgroup analysis revealed that individuals under the age of 65 were more vulnerable to the detrimental health impacts of PM2.5 constituents. Nevertheless, there remained a lack of unanimity regarding the influence of age on modifying the health risks associated with particulate matter exposure (Liu et al., 2021; Qi et al., 2024). For instance, in an analysis of the associations between PM2.5 constituents and cardiovascular diseases in China, Liu et al. (2021) discovered that the elderly exhibited greater sensitivity to such effects. Conversely, another study in China, carried out by Qi et al. (2024) which explored the association between PM2.5 constituents and cognitive function, indicated that young individuals demonstrated stronger susceptibility. These discrepant outcomes can be ascribed to factors such as geographical location, pollution concentrations, differences in pollutant composition, along with other latent health factors relevant to the specific group of people (Mokoena et al., 2019). In particular, factors such as increased outdoor activities and varied lifestyle leading to higher exposure to PM2.5 constituents among younger individuals (Zhou et al., 2022). As an illustration, in this study, we discovered that 72.55 % of young people and 61.71 % of older individuals engaged in regular exercise. The disparity between the two groups were statistically significant (P < 0.001). Moreover, given that elderly individuals have a diminished lung capacity, younger people might potentially inhale a greater quantity of environmental pollutants (Enright et al., 1994). This could also constitute a possible underlying reason. Additionally, we observed that married individuals demonstrated greater tolerance to the adverse health effects of black carbon, organic matter, and other PM2.5 constituents. This could be ascribed to the presence of social support and a more structured lifestyle among married individuals, which might help reduce stress levels and promote positive overall health (Manvelian and Sbarra, 2020). Specifically, married couples typically pooled their economic resources, affording them enhanced living conditions and consequently facilitating access to facilities and amenities that served to alleviate the impact of environmental hazards (Cohen and Wills, 1985). Moreover, compared to their unmarried counterparts, married individuals generally possess more expansive social networks (Manvelian and Sbarra, 2020). When confronted with health problems related to PM2.5 exposure, these supplementary connections can offer tangible forms of assistance. This might involve providing care and support during periods of illness or disseminating valuable information regarding preventive strategies (Tucker, 2002). The emotional sustenance derived from this broader social nexus further cushions the stress that accompanies potential health risks, bolstering the resilience of married individuals (Cohen and Wills, 1985). Moreover, our findings indicated that individuals with medical insurance for urban workers were more sensitive to the health hazards associated with black carbon. This could be linked to their higher exposure to road traffic pollution during daily commutes (Setton et al., 2008), as black carbon is a major constituent of vehicular emissions. Another potential rationale is that the distribution and sources of PM2.5 constituents varied across urban and rural areas, with urban areas often experiencing higher levels of pollution compared to rural regions (Yang et al., 2019, 2020). Therefore, we suggested that cities should pay more attention to pollution emissions from transportation sources (Jiang et al., 2017), while rural areas may need to consider more pollution from the burning of crop residues (Wu et al., 2020b). Conversely, participants with medical insurance for non-working urban residents were more vulnerable to other PM2.5 constituents. This might be attributed to individuals with medical insurance for non-working urban residents may encounter challenges in accessing adequate medical resources compared to those with insurance for urban workers (Li et al., 2018). Evidently, the current study revealed that the medical insurance for urban workers enjoyed a high level of financing and a diverse range of treatment options, whereas the comparatively lower-cost insurance for non-working urban residents and insurance for new rural cooperative offered more limited treatment (Wang et al., 2024b; Fan et al., 2021). Thus, the construction of more basic medical service facilities in rural areas and the optimization of medical services for non-employed people in cities should also be taken into account (Fan et al., 2021).

When results were stratified according to lifestyle indicators, we observed effect modification for alcohol consumption and exercise. Individuals who consumed alcohol exhibited greater tolerance to the risk of death associated with PM2.5 constituents exposure, a result consistent with prior studies (Yao et al., 2022). We hypothesized that individuals with poorer health conditions might tend to abstain from alcohol, and therefore, non-alcohol consumption might serve as a marker of poor health conditions and a higher risk of mortality, rather than being a causal factor for vulnerability to PM2.5 constituents exposure. While this finding may be the result of selection bias, additional studies are needed to support our hypothesis. Additionally, the modification effect of exercise observed in this study was complex, as exercise could lead to increased inhalation of PM2.5, resulting in higher health risks. However, exercise also provides overall health benefits (Tainio et al., 2021). Additional research is needed to have a more profound comprehension of the connection between exercise and the effects of PM2.5 constituents on health.

This study has several notable strengths. First, this is the initial study, to our knowledge, to explore the potential causal association between PM2.5 constituents and all-cause mortality using causal inference methods. This approach provides valuable insights into the potential causal links between these factors. Second, the large population size of the community survey ensured high statistical power and thereby greater generalizability of findings for similar populations.

Some limitations should be noted. First, the assumption of no unmeasured confounding factors, a prerequisite for causal inference, is challenging to meet in practice. Nonetheless, the sensitivity analysis conducted in this study, particularly the E-value analysis, indicated that the results remained robust even when considering potential unmeasured confounders. Second, the exposure assessment of PM2.5 constituents was based on residential addresses rather than individual-level data. This approach introduced the possibility of exposure misclassification, as individuals living in the same street were allocated the same exposure values. However, in reality, even when people are close geographically, their actual exposures vary due to their unique activity patterns. This type of error, known as a combination of Berkson error and classical error, might result in bias towards the null and an underestimation of the true effect (Bateson et al., 2007; Zeger et al., 2000). Finally, the pollutants concentrations in this study did not conform to a normal distribution, even after transformations, but we still used Gaussian distribution as the link function in propensity score. Gaussian distribution is still the optimal option based on the existing statistical theory and the limited number of link functions available. In this context, we have developed multiple models for estimating the generalized propensity score including a machine learning algorithm (XGboost) which does not depend on the distribution of pollutant concentrations. Moreover, based on the AC values, the generalized propensity score in this study can effectively balance the covariates, even when the pollutant concentrations did not follow a normal distribution. The same framework of methodology has been commonly used in previous studies (Wang et al., 2023; Robins et al., 2000; Zhang et al., 2024).

5. Conclusion

In conclusion, this large population-based cohort study in southern China established an independent potential causal relationship between each PM2.5 constituents and all-cause mortality. The findings highlighted the diverse health risks associated with different constituents, owing to variations in their sources and distribution patterns. Furthermore, the study demonstrated that individuals under 65 years old, unmarried, urban medical insured, not consuming alcohol, or not exercising exhibited a more pronounced relationship between PM2.5 constituents and all-cause mortality. Therefore, it is recommended that pollution control policies better shift their focus from PM2.5 mass to chemical composition, recognizing the importance of specific constituents. Furthermore, environmental health policies should prioritize the needs of vulnerable populations.

Supplementary Material

Supplemental Materials

Funding and Acknowledgments

This work was supported by National Natural Science Foundation of China (82204162, 81973150), Young Elite Scientist Sponsorship Program by China Association for Science and Technology (2023QNRC001), Guangdong Provincial Pearl River Talents Program (0920220207), Basic and Applied Basic Research Foundation of Guangdong Province (2022A1515010823, 2022A1515110263), Guangzhou Municipal Science and Technology Bureau (2023A04J2072, 202201011143), and Fundamental Research Funds for the Central Universities, Sun Yat-sen University (23qnpy108). Professor Yuantao Hao gratefully acknowledges the support of K.C.Wong Education Foundation

Footnotes

CRediT authorship contribution statement

Huanle Cai: Conceptualisation, Data curation, Formal analysis, Methodology, Software, Validation, Visualization, Writing—original draft, Writing–review and editing. Yuqin Zhang: Conceptualisation, Data curation, Methodology, Validation, Writing–review and editing. Jing Wei: Data curation, Methodology, Resources, Validation. Wayne R. Lawrence: Writing–review and editing. Shirui Chen: Conceptualisation, Data curation, Methodology. Huanhuan Cheng: Writing—review and editing. Hui Tang: Writing—review and editing. Dan Chen: Writing—review and editing. Ying Wang: Writing—review and editing. Jing Gu: Conceptualisation, Data curation, Funding acquisition, Project administration, Resources, Supervision, Validation, Writing—review and editing. Wangjian Zhang: Conceptualisation, Data curation, Funding acquisition, Project administration, Resources, Supervision, Validation, Writing—review and editing. Yuantao Hao: Conceptualisation, Data curation, Funding acquisition, Project administration, Resources, Supervision, Validation, Writing—review and editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.ecoenv.2025.117897.

Data availability

Data will be made available on request.

References

  1. Austin PC, 2019. Assessing covariate balance when using the generalized propensity score with quantitative or continuous exposures. Stat. Methods Med. Res 28 (5), 1365–1377. 10.1177/0962280218756159. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bateson TF, Coull BA, Hubbell B, Ito K, Jerrett M, Lumley T, Thomas D, Vedal S, Ross M, 2007. Panel discussion review: session three-issues involved in interpretation of epidemiologic analyses–statistical modeling. J. Expo. Sci. Environ. Epidemiol 17 (2), S90–S96. 10.1038/sj.jes.7500631. [DOI] [PubMed] [Google Scholar]
  3. Chen Y, Chen R, Chen Y, Dong X, Zhu J, Liu C, van Donkelaar A, Martin RV, Li H, Kan H, Jiang Q, Fu C, 2021. The prospective effects of long-term exposure to ambient PM2.5 and constituents on mortality in rural east china. Chemosphere 280, 130740. 10.1016/j.chemosphere.2021.130740. [DOI] [PubMed] [Google Scholar]
  4. Chen J, Hoek G, 2020. Long-term exposure to pm and all-cause and cause-specific mortality: a systematic review and meta-analysis. Environ. Int 143, 105974. 10.1016/j.envint.2020.105974. [DOI] [PubMed] [Google Scholar]
  5. Chen R, Kan H, Chen B, Huang W, Bai Z, Song G, Pan G, 2012. Association of particulate air pollution with daily mortality: the china air pollution and health effects study. Am. J. Epidemiol 175 (11), 1173–1181. 10.1093/aje/kwr425. [DOI] [PubMed] [Google Scholar]
  6. Chen S, Liu D, Huang L, Guo C, Gao X, Xu Z, Yang Z, Chen Y, Li M, Yang J, 2024. Global associations between long-term exposure to PM2.5 constituents and health: a systematic review and meta-analysis of cohort studies. J. Hazard. Mater 474, 134715. 10.1016/j.jhazmat.2024.134715. [DOI] [PubMed] [Google Scholar]
  7. Chen C, Xu D, He MZ, Wang Y, Du Z, Du Y, Qian Y, Ji D, Li T, 2018. Fine particle constituents and mortality: a time-series study in beijing, china. Environ. Sci. Technol 52 (19), 11378–11386. 10.1021/acs.est.8b00424. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cohen S, Wills TA, 1985. Stress, social support, and the buffering hypothesis. Psychol. Bull 98 (2), 310–357. [PubMed] [Google Scholar]
  9. Cole SR, Hernan MA, 2008. Constructing inverse probability weights for marginal structural models. Am. J. Epidemiol 168 (6), 656–664. 10.1093/aje/kwn164. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Collaborators GRF, 2020. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the global burden of disease study 2019. Lancet 396 (10258), 1223–1249. 10.1016/S0140-6736(20)30752-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Dominici F, Zigler C, 2017. Best practices for gauging evidence of causality in air pollution epidemiology. Am. J. Epidemiol 186 (12), 1303–1309. 10.1093/aje/kwx307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Enright PL, Kronmal RA, Manolio TA, Schenker MB, Hyatt RE, 1994. Respiratory muscle strength in the elderly. Correlates and reference values. Cardiovascular health study research group. Am. J. Respir. Crit. Care Med 149 (2 Pt 1), 430–438. 10.1164/ajrccm.149.2.8306041. [DOI] [PubMed] [Google Scholar]
  13. Fan X, Su M, Si Y, Zhao Y, Zhou Z, 2021. The benefits of an integrated social medical insurance for health services utilization in rural china: evidence from the china health and retirement longitudinal study. Int. J. Equity Health 20 (1), 126. 10.1186/s12939-021-01457-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Feng S, Gao D, Liao F, Zhou F, Wang X, 2016. The health effects of ambient PM2.5 and potential mechanisms. Ecotoxicol. Environ. Saf 128, 67–74. 10.1016/j.ecoenv.2016.01.030. [DOI] [PubMed] [Google Scholar]
  15. Fu J, Fei F, Wang S, Zhao Q, Yang X, Zhong J, Hu K, 2023. Short-term effects of fine particulate matter constituents on mortality considering the mortality displacement in zhejiang province, china. J. Hazard. Mater 457, 131723. 10.1016/j.jhazmat.2023.131723. [DOI] [PubMed] [Google Scholar]
  16. Goobie GC, Carlsten C, Johannson KA, Khalil N, Marcoux V, Assayag D, Manganas H, Fisher JH, Kolb M, Lindell KO, Fabisiak JP, Chen X, Gibson KF, Zhang Y, Kass DJ, Ryerson CJ, Nouraie SM, 2022. Association of particulate matter exposure with lung function and mortality among patients with fibrotic interstitial lung disease. JAMA Intern. Med 182 (12), 1248–1259. 10.1001/jamainternmed.2022.4696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Gruber S, Phillips RV, Lee H, van der Laan MJ, 2022. Data-adaptive selection of the propensity score truncation level for inverse-probability-weighted and targeted maximum likelihood estimators of marginal point treatment effects. Am. J. Epidemiol 191 (9), 1640–1651. 10.1093/aje/kwac087. [DOI] [PubMed] [Google Scholar]
  18. Haneuse S, Vanderweele TJ, Arterburn D, 2019. Using the e-value to assess the potential effect of unmeasured confounding in observational studies. JAMA 321 (6), 602–603. 10.1001/jama.2018.21554. [DOI] [PubMed] [Google Scholar]
  19. Hao H, Wang Y, Zhu Q, Zhang H, Rosenberg A, Schwartz J, Amini H, van Donkelaar A, Martin R, Liu P, Weber R, Russel A, Yitshak-Sade M, Chang H, Shi L, 2023. National cohort study of long-term exposure to PM2.5 components and mortality in medicare american older adults. Environ. Sci. Technol 57 (17), 6835–6843. 10.1021/acs.est.2c07064. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. He G, Wang Y, Cheng C, Guo J, Lin Z, Liang Z, Jin B, Tao L, Rong L, Chen L, Lin T, Hua Y, Park S, Mo Y, Li J, Jiang X, 2024. PM2.5 constituents associated with mortality and kidney failure in childhood-onset lupus nephritis: a 19-year cohort study. Sci. Total Environ 949, 175333. 10.1016/j.scitotenv.2024.175333. [DOI] [PubMed] [Google Scholar]
  21. Hvidtfeldt UA, Geels C, Sorensen M, Ketzel M, Khan J, Tjonneland A, Christensen JH, Brandt J, Raaschou-Nielsen O, 2019. Long-term residential exposure to pm(2.5) constituents and mortality in a danish cohort. Environ. Int 133 (Pt B), 105268. 10.1016/j.envint.2019.105268. [DOI] [PubMed] [Google Scholar]
  22. Janes H, Dominici F, Zeger S, 2010. On quantifying the magnitude of confounding. Biostatistics 11 (3), 572–582. 10.1093/biostatistics/kxq007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Jiang B, Liang S, Peng ZR, Cong H, Levy M, Cheng Q, Wang T, Remais JV, 2017. Transport and public health in china: the road to a healthy future. Lancet 390 (10104), 1781–1791. 10.1016/S0140-6736(17)31958-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kang N, Wu R, Liao W, Zhang C, Liu X, Mao Z, Huo W, Hou J, Zhang K, Tian H, Lin H, Wang C, 2023. Association of long-term exposure to PM2.5 constituents with glucose metabolism in chinese rural population. Sci. Total Environ 859 (Pt 2), 160364. 10.1016/j.scitotenv.2022.160364. [DOI] [PubMed] [Google Scholar]
  25. Li W, Wilker EH, Dorans KS, Rice MB, Schwartz J, Coull BA, Koutrakis P, Gold DR, Keaney JJ, Lin H, Vasan RS, Benjamin EJ, Mittleman MA, 2016. Short-term exposure to air pollution and biomarkers of oxidative stress: the framingham heart study. J. Am. Heart Assoc 5 (5). 10.1161/JAHA.115.002742. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Li T, Zhang Y, Wang J, Xu D, Yin Z, Chen H, Lv Y, Luo J, Zeng Y, Liu Y, Kinney PL, Shi X, 2018. All-cause mortality risk associated with long-term exposure to ambient PM2.5 in china: a cohort study. Lancet Public Health 3 (10), e470–e477. 10.1016/S2468-2667(18)30144-0. [DOI] [PubMed] [Google Scholar]
  27. Lipsky AM, Greenland S, 2022. Causal directed acyclic graphs. JAMA 327 (11), 1083–1084. 10.1001/jama.2022.1816. [DOI] [PubMed] [Google Scholar]
  28. Liu C, Chen R, Sera F, Vicedo-Cabrera AM, Guo Y, Tong S, Coelho M, Saldiva P, Lavigne E, Matus P, Valdes ON, Osorio GS, Pascal M, Stafoggia M, Scortichini M, Hashizume M, Honda Y, Hurtado-Diaz M, Cruz J, Nunes B, Teixeira JP, Kim H, Tobias A, Iniguez C, Forsberg B, Astrom C, Ragettli MS, Guo YL, Chen BY, Bell ML, Wright CY, Scovronick N, Garland RM, Milojevic A, Kysely J, Urban A, Orru H, Indermitte E, Jaakkola J, Ryti N, Katsouyanni K, Analitis A, Zanobetti A, Schwartz J, Chen J, Wu T, Cohen A, Gasparrini A, Kan H, 2019. Ambient particulate air pollution and daily mortality in 652 cities. N. Engl. J. Med 381 (8), 705–715. 10.1056/NEJMoa1817364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Liu D, Li J, Zhang Y, Xu Y, Liu X, Ding P, Shen C, Chen Y, Tian C, Zhang G, 2013. The use of levoglucosan and radiocarbon for source apportionment of PM2.5 carbonaceous aerosols at a background site in east china. Environ. Sci. Technol 47 (18), 10454–10461. 10.1021/es401250k. [DOI] [PubMed] [Google Scholar]
  30. Liu L, Zhang Y, Yang Z, Luo S, Zhang Y, 2021. Long-term exposure to fine particulate constituents and cardiovascular diseases in chinese adults. J. Hazard. Mater 416, 126051. 10.1016/j.jhazmat.2021.126051. [DOI] [PubMed] [Google Scholar]
  31. Manvelian A, Sbarra DA, 2020. Marital status, close relationships, and all-cause mortality: results from a 10-year study of nationally representative older adults. Psychosom. Med 82 (4), 384–392. 10.1097/PSY.0000000000000798. [DOI] [PubMed] [Google Scholar]
  32. Mokoena KK, Ethan CJ, Yu Y, Shale K, Liu F, 2019. Ambient air pollution and respiratory mortality in xi’an, china: a time-series analysis. Respir. Res 20 (1), 139. 10.1186/s12931-019-1117-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Mukherjee A, Agrawal M, 2018. A global perspective of fine particulate matter pollution and its health effects. Rev. Environ. Contam. Toxicol 244, 5–51. 10.1007/398_2017_3. [DOI] [PubMed] [Google Scholar]
  34. Niranjan R, Thakur AK, 2017. The toxicological mechanisms of environmental soot (black carbon) and carbon black: focus on oxidative stress and inflammatory pathways. Front. Immunol 8, 763. 10.3389/fimmu.2017.00763. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Niu Y, Chen R, Xia Y, Cai J, Ying Z, Lin Z, Liu C, Chen C, Peng L, Zhao Z, Zhou W, Chen J, Wang D, Huo J, Wang X, Fu Q, Kan H, 2018. Fine particulate matter constituents and stress hormones in the hypothalamus-pituitary-adrenal axis. Environ. Int 119, 186–192. 10.1016/j.envint.2018.06.027. [DOI] [PubMed] [Google Scholar]
  36. Orellano P, Reynoso J, Quaranta N, Bardach A, Ciapponi A, 2020. Short-term exposure to particulate matter (PM10 and PM2.5), nitrogen dioxide (NO2), and ozone (O3) and all-cause and cause-specific mortality: systematic review and meta-analysis. Environ. Int 142, 105876. 10.1016/j.envint.2020.105876. [DOI] [PubMed] [Google Scholar]
  37. Ostro B, Lipsett M, Reynolds P, Goldberg D, Hertz A, Garcia C, Henderson KD, Bernstein L, 2010. Long-term exposure to constituents of fine particulate air pollution and mortality: results from the california teachers study. Environ. Health Perspect 118 (3), 363–369. 10.1289/ehp.0901181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Pateraki S, Asimakopoulos DN, Maggos T, Assimakopoulos VD, Bougiatioti A, Bairachtari K, Vasilakos C, Mihalopoulos N, 2020. Chemical characterization, sources and potential health risk of PM2.5 and PM1 pollution across the greater athens area. Chemosphere 241, 125026. 10.1016/j.chemosphere.2019.125026. [DOI] [PubMed] [Google Scholar]
  39. Qi J, Zhao N, Liu M, Guo Y, Fu J, Zhang Y, Wang W, Su Z, Zeng Y, Yao Y, Hu K, 2024. Long-term exposure to fine particulate matter constituents and cognitive impairment among older adults: an 18-year chinese nationwide cohort study. J. Hazard. Mater 468, 133785. 10.1016/j.jhazmat.2024.133785. [DOI] [PubMed] [Google Scholar]
  40. Robins JM, Hernan MA, Brumback B, 2000. Marginal structural models and causal inference in epidemiology. Epidemiology 11 (5), 550–560. 10.1097/00001648-200009000-00011. [DOI] [PubMed] [Google Scholar]
  41. Ruan B, Yu Z, Yang S, Xu K, Ren J, Yao J, Wu N, Yu C, Deng M, Xie T, Chen P, Wang C, Li Y, Zhao Y, Sheng J, Hou Y, Wu Z, Jin S, Chen Y, Li M, Zhu F, Tang H, Hao Y, Pang X, Lu L, Yang W, Yuan Z, Xu A, Li Z, Ni M, Yan Y, Zhong Q, Zhou L, Li G, Meng Q, Hu J, Zhou H, Zhang G, Li D, Jiang W, Li Q, Wu P, Xing R, Gu J, Gao D, Li L, 2019. Establishment and development of national community-based collaborative innovation demonstration areas to achieve the control target of hepatitis b in china. BMC Infect. Dis 19 (1), 617. 10.1186/s12879-019-4150-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Selvin S, {Statistical analysis of epidemiologic data}, Oxford University Press, 2004. [Google Scholar]
  43. Setton EM, Keller CP, Cloutier-Fisher D, Hystad PW, 2008. Spatial variations in estimated chronic exposure to traffic-related air pollution in working populations: a simulation. Int. J. Health Geogr 7, 39. 10.1186/1476-072X-7-39. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Shi L, Zhu Q, Wang Y, Hao H, Zhang H, Schwartz J, Amini H, van Donkelaar A, Martin RV, Steenland K, Sarnat JA, Caudle WM, Ma T, Li H, Chang HH, Liu JZ, Wingo T, Mao X, Russell AG, Weber RJ, Liu P, 2023. Incident dementia and long-term exposure to constituents of fine particle air pollution: a national cohort study in the united states. Proc. Natl. Acad. Sci. USA 120 (1), e2083685177. 10.1073/pnas.2211282119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Tainio M, Jovanovic AZ, Nieuwenhuijsen MJ, Hu L, de Nazelle A, An R, Garcia L, Goenka S, Zapata-Diomedi B, Bull F, 2021. T.H. Sa, Air pollution, physical activity and health: a mapping review of the evidence. Environ. Int 147, 105954. 10.1016/j.envint.2020.105954. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Tao B, Li Z, Wang Y, Wu J, Shi X, Shi J, Liu Q, Wang J, 2023. Environment pollutants exposure affects the endogenous activation of within-host mycobacterium tuberculosis. Environ. Res 227, 115695. 10.1016/j.envres.2023.115695. [DOI] [PubMed] [Google Scholar]
  47. Tucker JS, 2002. Health-related social control within older adults’ relationships. J. Gerontol. Ser. B-Psychol. Sci. Soc. Sci 57 (5), P387–P395. 10.1093/geronb/57.5.p387. [DOI] [PubMed] [Google Scholar]
  48. Wang Y, Du Z, Zhang W, Wang X, Lin X, Liu Y, Deng Y, Zhang D, Gu J, Xu L, Hao Y, 2024a. Cohort profile: the pearl river cohort study. Int. J. Epidemiol 53 (5). 10.1093/ije/dyae112. [DOI] [PubMed] [Google Scholar]
  49. Wang X, Guo Y, Cai M, Qian ZM, Zhang S, Zhang Z, Yang Y, Vaughn MG, Aaron HE, Wu F, Zhang Y, Lin H, 2022. Constituents of fine particulate matter and asthma in 6 low- and middle-income countries. J. Allergy Clin. Immunol 150 (1), 214–222. 10.1016/j.jaci.2021.12.779. [DOI] [PubMed] [Google Scholar]
  50. Wang Y, Shi Z, Shen F, Sun J, Huang L, Zhang H, Chen C, Li T, Hu J, 2019. Associations of daily mortality with short-term exposure to PM2.5 and its constituents in shanghai, china. Chemosphere 233, 879–887. 10.1016/j.chemosphere.2019.05.249. [DOI] [PubMed] [Google Scholar]
  51. Wang Y, Wei J, Zhang Y, Guo T, Chen S, Wu W, Chen S, Li Z, Qu Y, Xiao J, Deng X, Liu Y, Du Z, Zhang W, Hao Y, 2023. Estimating causal links of long-term exposure to particulate matters with all-cause mortality in south china. Environ. Int 171, 107726. 10.1016/j.envint.2022.107726. [DOI] [PubMed] [Google Scholar]
  52. Wang E, Zhu M, Lin Y, Xi X, 2024b. Multilevel medical insurance mitigate health cost inequality due to air pollution: evidence from china. Int. J. Equity Health 23 (1), 153. 10.1186/s12939-024-02238-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Wei J, Li Z, Chen X, Li C, Sun Y, Wang J, Lyapustin A, Brasseur GP, Jiang M, Sun L, Wang T, Jung CH, Qiu B, Fang C, Liu X, Hao J, Wang Y, Zhan M, Song X, Liu Y, 2023. Separating daily 1 km PM2.5 inorganic chemical composition in china since 2000 via deep learning integrating ground, satellite, and model data. Environ. Sci. Technol 10.1021/acs.est.3c00272. [DOI] [PubMed] [Google Scholar]
  54. Wei J, Li Z, Cribb M, Huang W, Xue W, Sun L, Guo J, Peng Y, Li J, Lyapustin A, Liu L, Wu H, Song Y, 2020. Improved 1km resolution PM2.5 estimates across china using enhanced space–time extremely randomized trees. Atmos. Chem. Phys 20 (6), 3273–3289. 10.5194/acp-20-3273-2020. [DOI] [Google Scholar]
  55. Wei J, Li Z, Lyapustin A, Sun L, Peng Y, Xue W, Su T, Cribb M, 2021. Reconstructing 1-km-resolution high-quality PM2.5 data records from 2000 to 2018 in china: spatiotemporal variations and policy implications. Remote Sens. Environ 252, 112136. 10.1016/j.rse.2020.112136. [DOI] [Google Scholar]
  56. Wu X, Braun D, Schwartz J, Kioumourtzoglou MA, Dominici F, 2020a. Evaluating the impact of long-term exposure to fine particulate matter on mortality among the elderly. Sci. Adv 6 (29), eaba5692. 10.1126/sciadv.aba5692. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Wu J, Kong S, Wu F, Cheng Y, Zheng S, Qin S, Liu X, Yan Q, Zheng H, Zheng M, Yan Y, Liu D, Ding S, Zhao D, Shen G, Zhao T, Qi S, 2020b. The moving of high emission for biomass burning in china: view from multi-year emission estimation and human-driven forces. Environ. Int 142, 105812. 10.1016/j.envint.2020.105812. [DOI] [PubMed] [Google Scholar]
  58. Yang Y, Ruan Z, Wang X, Yang Y, Mason TG, Lin H, Tian L, 2019. Short-term and long-term exposures to fine particulate matter constituents and health: a systematic review and meta-analysis. Environ. Pollut 247, 874–882. 10.1016/j.envpol.2018.12.060. [DOI] [PubMed] [Google Scholar]
  59. Yang J, Zhou M, Li M, Yin P, Hu J, Zhang C, Wang H, Liu Q, Wang B, 2020. Fine particulate matter constituents and cause-specific mortality in china: a nationwide modelling study. Environ. Int 143, 105927. 10.1016/j.envint.2020.105927. [DOI] [PubMed] [Google Scholar]
  60. Yao Y, Lu T, Liu Y, Qin Q, Jiang J, Xiang H, 2022. Association of depressive symptoms with ambient pm(2.5) in middle-aged and elderly chinese adults: a cross-sectional study from the china health and retirement longitudinal study wave 4. Environ. Res 203, 111889. 10.1016/j.envres.2021.111889. [DOI] [PubMed] [Google Scholar]
  61. Zeger SL, Thomas D, Dominici F, Samet JM, Schwartz J, Dockery D, Cohen A, 2000. Exposure measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health Perspect 108 (5), 419–426. 10.1289/ehp.00108419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Zhang Z, 2016. Multiple imputation with multivariate imputation by chained equation (mice) package. Ann. Transl. Med 4 (2), 30. 10.3978/j.issn.2305-5839.2015.12.63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Zhang Y, Chen S, Wei J, Jiang J, Lin X, Wang Y, Hao C, Wu W, Yuan Z, Sun J, Wang H, Du Z, Zhang W, Hao Y, 2024. Long-term PM1 exposure and hypertension hospitalization: a causal inference study on a large community-based cohort in south china. Sci. Bull 69 (9), 1313–1322. 10.1016/j.scib.2024.03.028. [DOI] [PubMed] [Google Scholar]
  64. Zhang J, Cheng H, Wang D, Zhu Y, Yang C, Shen Y, Yu J, Li Y, Xu S, Zhang S, Song X, Zhou Y, Chen J, Jiang J, Fan L, Wang C, Hao K, 2021. Chronic exposure to PM2.5 nitrate, sulfate, and ammonium causes respiratory system impairments in mice. Environ. Sci. Technol 55 (5), 3081–3090. 10.1021/acs.est.0c05814. [DOI] [PubMed] [Google Scholar]
  65. Zhang Z, Reinikainen J, Adeleke KA, Pieterse ME, Groothuis-Oudshoorn C, 2018. Time-varying covariates and coefficients in cox regression models. Ann. Transl. Med 6 (7), 121. 10.21037/atm.2018.02.12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Zhou P, Hu J, Yu C, Bao J, Luo S, Shi Z, Yuan Y, Mo S, Yin Z, Zhang Y, 2022. Short-term exposure to fine particulate matter constituents and mortality: case-crossover evidence from 32 counties in china. Sci. China Life Sci 65 (12), 2527–2538. 10.1007/s11427-021-2098-7. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Materials
NIHMS2058416-supplement-Supplemental_Materials.docx (753.5KB, docx)

Data Availability Statement

Data will be made available on request.

RESOURCES