Assessing the timing and the duration of exposure to air pollution on cardiometabolic biomarkers in patients suspected of coronary artery disease

Abstract

Air pollution can affect cardiometabolic biomarkers in susceptible populations, but the most important exposure window (lag days) and exposure duration (length of averaging period) are not well understood. We investigated air pollution exposure across different time intervals on ten cardiometabolic biomarkers in 1550 patients suspected of coronary artery disease. Daily residential PM_2.5 and NO₂ were estimated using satellite-based spatiotemporal models and assigned to participants for up to one year before the blood collection. Distributed lag models and generalized linear models were used to examine the single-day-effects by variable lags and cumulative effects of exposures averaged over different periods before the blood draw. In single-day-effect models, PM_2.5 was associated with lower apolipoprotein A (ApoA) in the first 22 lag days with the effect peaking on the first lag day; PM_2.5 was also associated with elevated high-sensitivity C-reactive protein (hs-CRP) with significant exposure windows observed after the first 5 lag days. For the cumulative effects, short- and medium-term exposure was associated with lower ApoA (up to 30wk-average) and higher hs-CRP (up to 8wk-average), triglycerides and glucose (up to 6d-average), but the associations were attenuated to null over the long term. The impacts of air pollution on inflammation, lipid, and glucose metabolism differ by the exposure timing and durations, which can inform our understanding of the cascade of underlying mechanisms among susceptible patients.

1. Introduction

Numerous studies have suggested associations of exposure to ambient air pollution with an increased risk of cardiometabolic diseases (Brook et al., 2010; Rajagopalan et al., 2018). Although the underlying pathophysiologic mechanisms are complex and remain inconclusive, some potential biological pathways have been illustrated (Araujo, 2011; Brook et al., 2010; Sun et al., 2009). Fine particulate matter (PM_2.5, aerodynamic diameter <2.5 μm) can local and systemic oxidative stress and inflammation, and disrupt the functions of multiple target cells in cardiovascular system, resulting in a series of pathological changes, such as endothelial dysfunction, atherosclerosis progression, platelet aggregation and thrombosis, abnormal lipid metabolism, impaired glucose metabolism, and insulin resistance, ultimately leading to cardiovascular events (Brook et al., 2010; Chin, 2015; Feng et al., 2023; Rao et al., 2014). Nitrogen dioxide (NO₂) is one of the highly reactive gaseous pollutants that can generate free radicals, induce inflammatory response, trigger lipid oxidation and impair lipid metabolism, and affect glucose metabolic homeostasis (Guo et al., 2023; Ji et al., 2015; Lucht et al., 2018). Air pollution-associated cardiometabolic diseases, including cardiovascular disease (CVD) and type 2 diabetes, usually take years to develop; some cardiometabolic biomarkers can serve as preclinical indicators for early identification of cardiometabolic health risk, particularly among susceptible people who are at higher risk for atherosclerosis. C-reactive protein (CRP) represents a sensitive circulating inflammatory biomarker that is consistently associated with cardiovascular risk (Araujo, 2011; Brook et al., 2010; Rifai and Ridker, 2001). Major blood lipids (e.g., triglycerides [TG], total cholesterol [TC], high-density lipoprotein cholesterol [HDL-C], low-density lipoprotein cholesterol [LDL-C]) and apolipoproteins have been proven to play an important role in the development of cardiovascular events (Abdullah et al., 2018; Boekholdt et al., 2012; Ridker et al., 2005). TG/HDL-C ratio and TyG index are surrogate markers of insulin resistance (Giannini et al., 2011; Simental-Mendía et al., 2008) and have been significantly associated with increased risk of diabetes and CVD incidence (Tian et al., 2021; Vega et al., 2014). Assessing how cardiometabolic biomarkers respond to air pollution exposure may provide insight into the disease risk detection and pathophysiological pathways by which air pollution contributes to adverse cardiometabolic outcomes.

The impact of air pollution on health outcomes is related to the timing of exposure. A number of epidemiological studies have confirmed the associations of both short- and long-term air pollution exposure with changes in cardiometabolic biomarkers, including CRP (Chen et al., 2017; Dadvand et al., 2014; Li et al., 2017; Ostro et al., 2014; Xu et al., 2019b; Xu et al., 2022b; Zhang et al., 2017), glucose (Chen et al., 2016b; Liu et al., 2019a; Lucht et al., 2018; Yitshak Sade et al., 2016; Zhan et al., 2021), and lipid profiles (Chen et al., 2016b; Gaio et al., 2019; McGuinn et al., 2019; Yitshak Sade et al., 2016). In the short term, the health effects may occur on the present day (no lag) or days following the exposure (lagged effects); there is likely a sensitive window in which air pollution exposure is more biologically relevant to the health outcomes. Results on how air pollution exposure in a single-day or moving average lags affected cardiometabolic biomarkers were mixed (Lee et al., 2018; Li et al., 2017; Yitshak Sade et al., 2016; Zhan et al., 2021), which might be because previous studies only examined lagged effects within a very short time period (e.g., 1 week) and did not systematically evaluate the appropriate exposure window. In addition, persistent exposure over time might result in cumulative effects that might not be fully captured if the exposure is only averaged in short periods (Pope III, 2007). Previous research on cumulative effects of air pollution exposure focused on pre-defined intervals to average exposures, ranging from several weeks to a few months (Bell et al., 2017; Chen et al., 2016b; Lucht et al., 2018; Tsai et al., 2019). Examining changes in biomarkers associated with exposure to air pollutants within different windows over a relatively longer period in the same study population may allow us to compare the magnitude of impacts over time. To date, however, studies addressing both short-term (exposure averaged within a few days and several weeks) and long-term (exposure averaged over 6 months) effects of PM_2.5 and NO₂ exposure on CRP (Dadvand et al., 2014; Green et al., 2015; Hajat et al., 2015; Iyer et al., 2022; Lee et al., 2018; Panasevich et al., 2009), lipid profile (Bell et al., 2017; Chen et al., 2016b; Wu et al., 2019; Xiao et al., 2016; Yitshak Sade et al., 2016), or glucose levels (Chen et al., 2016b; Lucht et al., 2018; Yitshak Sade et al., 2016) have yielded inconsistent results. Besides, most studies have been focused on the generally healthy population, while research in susceptible populations is limited.

In this study, we aimed to examine the short-term lagged effects and longer-term cumulative effects of PM_2.5 and NO₂ on ten cardiometabolic biomarkers in a patient population with suspected coronary artery disease (CAD). Our analysis used flexible time intervals to explore the important exposure window and appropriate exposure duration.

2. Materials and Methods

2.1. Study Population

The study population is a subset of participants in the CREATION cohort, an ongoing prospective cohort in China. The study design and participant characteristics have been previously described (Hou et al., 2020; Hu et al., 2022; Wang et al., 2019). In brief, outpatients of adults were recruited at Fuwai Hospital in Beijing, China between 2015 and 2017. Eligible patients included patients who were 25 to 92 years, referred by their cardiologists over concern for CAD based on the American College of Cardiology (ACC)/American Heart Association (AHA) guidelines for atherosclerotic cardiovascular diseases (ASCVD) risk, and underwent cardiac computed tomography. Patients with a history of coronary events (coronary revascularization, myocardial infarction, and other heart diseases) were excluded. Over 92% of included patients had an ASCVD risk score <20%, representing a cohort of low-to-moderate risk patients. In-person interviews, blood sample collection, and clinical examinations were conducted according to standard protocols at baseline. All participants provided written informed consent, and the study was approved by the Institutional Review Boards of the Chinese Academy of Medical Sciences Fuwai Hospital and the University at Buffalo.

In the present study, we included 1550 participants who resided in the Beijing metropolitan area and had available blood samples at baseline. The analysis was restricted to local residents in Beijing to minimize the potential impact of selection bias due to differences in the recruitment process across regions. The characteristics of the study population are summarized in Table 1.

Table 1.

Characteristics of study participants (n=1550) and biomarkers

Characteristics	Mean ± SD or n (%)
Age (year)	59.5 ± 11.2
Male	787 (50.8)
Education
<College	1319 (85.4)
College	174 (11.3)
>College	51 (3.3)
Body mass index (kg/m2)	25.3 ± 3.3
<25	752 (48.5)
25–30	664 (42.8)
≥30	134 (8.7)
Physical activity (%)
Never	237 (15.6)
≤3 times per week	779 (51.1)
>3	508 (33.3)
Current smokers	412 (26.6)
Smoking years (in smokers)	31.0 ± 10.8
Cigarettes per day (in smokers)
≤10	65 (15.8)
10–20	252 (61.2)
>20	95 (23.1)
Alcohol use	287 (18.6)
Antihypertensive medication	727 (46.9)
Statin medication	519 (33.5)
Urban area	811 (52.3)
Payment
Fully self-paid	159 (10.3)
Partially self-paid	1325 (85.5)
Fully-covered	66 (4.3)
Coronary artery calcium score >0	611 (39.4)
Biomarkers	Median (Q1, Q3)
Glucose (mmol/L)	5.5 (5.0, 6.2)
Total cholesterol (mmol/L)	4.8 (4.0, 5.5)
Triglyceride (mmol/L)	1.4 (1.0, 2.0)
HDL-C (mmol/L)	1.3 (1.1, 1.5)
LDL-C (mmol/L)	3.0 (2.3, 3.6)
hs-CRP (mg/L)	1.4 (0.8, 2.5)
TG/HDL-C ratio	1.1 (0.7, 1.7)
Apolipoprotein A (g/L)	1.5 (1.3, 1.7)
Apolipoprotein B (g/L)	0.9 (0.8, 1.1)
TyG	1.4 (1.0, 1.8)

HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol; hs-CRP: high-sensitivity C-reactive protein; TyG: triglyceride-glucose. Normal range of biomarkers: glucose: 3.58–6.05 mmol/L; total cholesterol: 3.64–5.98 mmol/L; triglyceride: 0.38–1.76 mmol/L; HDL-C: 0.70–1.59 mmol/L; LDL-C: <3.37 mmol/L in the general population and <2.59 mmol/L in the high-risk group; hs-CRP: 0.00–3.00 mg/L; Apolipoprotein A: 1.1–1.8 g/L; Apolipoprotein B: 0.5–1.2 g/L.

2.2. Air Pollution Assessment

The daily concentrations of PM_2.5 and NO₂ were estimated at each participant’s geocoded residential address using a hierarchical machine-learning based modelling approach with a high spatial resolution (within the 6th ring road: 100m; outside: 1km). Detailed information on the model development and performance has been described previously (Huang et al., 2021; Huang et al., 2022; Xu et al., 2019a). Briefly, the prediction models incorporated daily mean ground-level PM_2.5 and NO₂ concentrations from regulatory monitors, extensive geographic predictors variables from multiple sources (e.g., traffic network, industrial emissions, population density, and land use), meteorological data, satellite-based air pollution data, and regression residuals smoothed by universal kriging. The model presented good performance with R² of 0.89 for PM_2.5 and 0.78 for NO₂ based on out-of-sample validation. In our current analysis, daily residential exposure to PM_2.5 and NO₂ was estimated and assigned to each participant for up to one year before the date of the blood draw. Figure 1 shows the design for determining the timing and durations of air pollution exposure.

Figure 1. Illustration of exposure windows and exposure duration defined in this study.

A single-day effect refers to the effect of air pollution exposure on a single day by multiple lag days (i.e., effects on the 1^st, 2^nd, …, up to 28^th day before the blood draw). A cumulative effect refers to the effect of air pollution exposure averaged over different durations since the blood draw (i.e., averaged from the 1^st day before the blood draw to the 2^nd [1–2d-average], the 3^rd [1–3d-average], …, the 364^th day [52 week-average] before the blood draw).

2.3. Blood Collection and Measurement

After overnight fasting for at least 8h, peripheral venous blood samples were collected by certified nurses using coagulant vacuum tubes between 6:00 AM and 10:00 AM. Serum concentrations of high-sensitivity C-reactive protein (hs-CRP) and blood lipid parameters (TC, TG, HDL-C, and LDL-C) and apolipoproteins (ApoA and ApoB) were measured by automatic analyzer (Beckman AU5800; Beckman Coulter) at the clinical laboratory of Fuwai Hospital. The ratio of triglycerides to high-density lipoprotein cholesterol [TG:HDL-C] was calculated. Fasting glucose was measured by a biochemistry analyzer (AU2700; Olympus, Tokyo, Japan). TyG index was the natural log-transformed product of (TG × glucose × 0.5) (Simental-Mendía et al., 2008).

2.4. Covariate Assessment

Questionnaires were administered by trained interviewers to collect information on demographics (e.g., age, sex, education), lifestyle habits (e.g., physical activity, smoking, alcohol use), and medical history (e.g., hypertension, diabetes, medication uses). Height and weight were measured at the baseline clinic visit by trained nurses. Body mass index (BMI) was calculated as weight divided by height squared (kg/m²). Coronary artery calcium (CAC) was measured by a second-generation dual-source CT system (SOMATOM Definition Flash; Siemens Healthcare, Forchheim, Germany) with a standardized scanning protocol (Hou et al., 2012). Coronary calcium lesions were defined as having a threshold ≥130 Hounsfield units (HUs) and an area ≥1 mm². The total calcium burden in the coronary arteries was quantified by the scoring algorithm proposed by Agatston et al (Agatston et al., 1990).

2.5. Statistical Analysis

Histograms and Q-Q plots were used to assess the normality of the data. Log-transformations were applied to all biomarkers to improve normality for further modeling. Characteristics of the study population were examined by summary statistics (e.g., mean, standard deviation [SD], frequency, percentage).

We used distributed lag non-linear models (DLNM) to examine the single-day effects of exposure to PM_2.5 and NO₂ by multiple lag days before the blood draw (lag 1 to lag 28, Figure 1) on the levels of each biomarker. The analyses were performed in R using the “dlnm” package (Gasparrini, 2011). DLNM is based on the definition of a “cross-basis”, a bi-dimensional space of functions to describe both the exposure-response relationship and lag structure of the association, accounting for the high correlation between air pollution levels on adjacent days (Gasparrini, 2014; Gasparrini et al., 2010). We first examined the exposure-response relationship using a natural cubic spline function with a 3-degree of freedom. As the results implied linearity, we then used a linear function for the exposure-response relationship. The non-linear effects across lags were modeled with a natural cubic spline function using a 3-degree of freedom. Finally, these functions were combined to simultaneously account for the lag effects and exposure-response relationship of air pollution exposure. A set of potential confounders were selected a priori and adjusted in the models, including age (continuous), sex (male, female), education (<college, college, >college), BMI (continuous), smoking status (never/past, current), smoking years (continuous), cigarettes per day (continuous), alcohol use (no, yes), physical activity (never, ≤3 times/wk, >3 times/wk), urbanization (urban, rural), payment method (fully self-paid, partially self-paid, fully-covered), antihypertensive medication use (no, yes), statin use (no, yes). We further adjusted for daily mean temperature and relative humidity on the current day using natural splines with a 3-degree of freedom to remove the effect of seasonality. The DLM is written as:

$$Y_i={\beta }_0+{\sum }_{j=1}^n\left[{\alpha }_j{\mathit{\text{Pollution}}}_{ij}\right]+{\beta }_1{\mathit{\text{Age}}}_i+{\beta }_2{\mathit{\text{Sex}}}_i+{\beta }_3{\mathit{\text{Education}}}_i+{\beta }_4{\mathit{\text{BMI}}}_i+{\beta }_5{\mathit{\text{Smoking}}}_i+{\beta }_6\mathit{\text{Smoking}}{\mathit{\text{ years}}}_i+{\beta }_7{\mathit{\text{Cigarettes per day}}}_i+{\beta }_8{\mathit{\text{Alcohol}}}_i+{\beta }_9{\mathit{\text{Exercise}}}_i+{\beta }_{10}{\mathit{\text{Urbanization}}}_i+{\beta }_{11}{\mathit{\text{Payment}}}_i+{\beta }_{12}{\mathit{\text{Hypertension medication}}}_i+{\beta }_{13}{\mathit{\text{Statin}}}_i+ns\left(\mathit{\text{Temp}},df\right)+ns\left(RH, df\right)+{\epsilon }_i$$

Where $Y_i$ refers to log-transformed levels of a specific biomarker for participant $i;{\alpha }_j$ are coefficients from a cross-basis function that combines dose-response relationship (linear) and lag-response relationship (a natural cubic spline function with a 3-degree of freedom); ${\mathit{\text{Pollution}}}_{ij}$ is the daily level of the air pollutant for participant $i$ at lag $j,{\beta }_1$, …, ${\beta }_{13}$ refer to the coefficients of aforementioned covariates, ns is the

We fit generalized linear models (GLM) to examine the cumulative effects of levels of PM_2.5 and NO₂ averaged over different periods before the blood draw on each biomarker. Daily exposure data were averaged by different durations from 2 days until 52 weeks (Figure 1). We modeled each biomarker with air pollution exposure averaged in each duration separately and each GLM is written as:

$$Y_i={\beta }_0+{\beta }_1{\mathit{\text{Pollutant}}}_i+{\beta }_2{\mathit{\text{Age}}}_i+{\beta }_3{\mathit{\text{Sex}}}_i+{\beta }_4{\mathit{\text{Education}}}_i+{\beta }_5{\mathit{\text{BMI}}}_i+{\beta }_6{\mathit{\text{Smoking}}}_i+{\beta }_7{\mathit{\text{Smoking years}}}_i+{\beta }_8{\mathit{\text{Cigarettes per day}}}_i+{\beta }_9{\mathit{\text{Alcohol}}}_i+{\beta }_{10}{\mathit{\text{Exercise}}}_i+{\beta }_{11}{\mathit{\text{Urbanization}}}_i+{\beta }_{12}{\mathit{\text{Payment}}}_i+{\beta }_{13}{\mathit{\text{Hypertension medication}}}_i+{\beta }_{14}{\mathit{\text{Statin}}}_i+ns\left({\mathit{\text{Temp}}}_i, df=3\right)+ns\left(R{H}_i, df=3\right)+{\epsilon }_i$$

Where $Y_i$ refers to log-transformed levels of a specific biomarker for participant $i$; ${\beta }_1$ refers to the coefficient for the air pollutant levels averaged over a specific duration for participant $i$, β₂, … β₁₄ refer to the coefficients of aforementioned covariates. Final results were reported as percent difference in levels of biomarkers for 10 μg/m³ increases in the concentrations of PM_2.5 and NO₂ as [exp (10×β) − 1] × 100% and their 95% CIs as (exp[10×(β±1.96×SE)]) – 1 × 100%, where β and SE are the regression coefficient of air pollutant levels averaged over a duration and its standard error, respectively.

For both single-day and cumulative effects, we further performed stratified analyses by sex (male, female), smoking (no, yes), BMI (<25, ≥25 kg/m²) and CAC score (0, >0). In the sensitivity analyses, we fit two-pollutant models to examine the independent effect of each pollutant on each biomarker. A z-test was conducted to compare the associations between air pollutants and biomarkers in subgroups: ${(\beta }_1-{\beta }_2)/\sqrt{{SE}_1^2+{SE}_1^2}$, where β₁ and β₂ are estimated effects and ${SE}_1$ and ${SE}_2$ are the standard errors.

All statistical analyses were conducted in R software version 4.1.2. All statistical tests were two-sided and p < 0.05 was considered as statistical significance.

3. Results

Of the 1550 participants with average age of 59. 5±11.2 years, 50.8% of the participants were males and 85.4% did not have a college degree (Table 1). The population characteristics of descriptive statistics of ten biomarkers are summarized in Table 1. The levels of biomarkers in the most of participants fall in the normal range. The variation in PM_2.5 and NO₂ and the correlations of each air pollutant during different periods decreased with longer averaging periods (Table S1 and Figure S1)

The results for single-day-effect models are presented in Figure 2 and Figure S2. NO₂ was positively associated with TC during the first 5 lag days and the effects peaked on the first lag day. Each 10 μg/m³ increase in NO₂ in the first and fifth lag days were associated with 0.23% (95% CI: 0.03, 0.43) and 0.12% (95% CI: 0.01, 0.22) increase in levels of TC, respectively. Similar patterns were observed for ApoB, with the greatest effects on the first lag day and then decreased. ApoA was negatively associated with PM_2.5 during the first 22 lag days and with NO₂ during the first 11 lag days, and the associations were strongest for the first day of exposure for both air pollutants. For hs-CRP and HDL-C, the significant single-day effects presented after first few lag days. For example, hs-CRP was positively associated with exposure to PM_2.5 starting from the lag 6 (0.23%, 95% CI: 0.01, 0.45) to the lag 23 days (0.21%, 95% CI: 0.00, 0.42) and with NO₂ during lag 4 (0.78%, 95% CI: 0.03, 1.53) to the lag 22 (0.51%, 95% CI: 0.01, 1.00). No associations were observed for glucose, TG, LDL-C, TG/HDL-C, or TyG index with any air pollutants, regardless the timing and duration of exposure.

Figure 2. The single-day effects of exposure to fine particulate matter (PM_2.5) and nitrogen dioxide (NO₂) by multiple lag days on levels of biomarkers using distributed lag models.

TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; ApoA, apolipoprotein A; ApoB, apolipoprotein B; hsCRP, high-sensitivity C-reactive protein. Models were adjusted for age (continuous), sex (male, female), education (<college, college, >college), body mass index (continuous), smoking status (never/past, current), smoking years (continuous), cigarettes per day (continuous), alcohol use (no, yes), physical activity (never, ≤3 times/wk, >3 times/wk), urbanization (urban, rural), payment method (fully self-paid, partially self-paid, fully-covered), antihypertensive medication use (no, yes), statin use (no, yes), temperature (df=3), and humidity (df=3). Dots represent percent difference percent difference in levels of biomarkers for 10 μg/m³ increases in PM_2.5 and NO₂ and error bars represent 95% confidence intervals. The red lines and error bars refer to significant associations and the black lines and error bars refer to non-significant associations.

In cumulative-effect models, the associations were significant and stronger in the shorter averaging periods (less than 6 months) compared to those in the longer term (over 6 months) (Figure 3 and Figure S3). For example, each 10 μg/m³ increase in PM_2.5 levels averaged over 2 weeks and over 30 weeks were associated with a decrease in ApoA by 0.54% (95% CI: −0.91, −0.17) and by 3.5% (95% CI: −5.98, −0.96), respectively, but associations became null after 30-week averages. Significant associations were presented among air pollution levels averaged in shorter periods for CRP (over 8-week average), glucose, TG, TG/HDL-C ratio, and TyG index (over one-week average). HDL-C was not significantly associated with NO₂ levels averaged over the first 10 weeks before the blood draw, but a positive association was found from 12-week to 46-week averages.

Figure 3. The cumulative effects of levels of fine particulate matter (PM_2.5) and nitrogen dioxide (NO₂) averaged over different periods before the blood draw on levels of biomarkers using generalized linear models.

TG, triglycerides; ApoA, apolipoprotein A; hsCRP, high- sensitivity C-reactive protein. Models were adjusted for age (continuous), sex (male, female), education (<college, college, >college), body mass index (continuous), smoking status (never/past, current), smoking years (continuous), cigarettes per day (continuous), alcohol use (no, yes), physical activity (never, ≤3 times/wk, >3 times/wk), urbanization (urban, rural), payment method (fully self-paid, partially self-paid, fully-covered), antihypertensive medication use (no, yes), statin use (no, yes), temperature (df=3), and humidity (df=3). Dots represent percent difference percent difference in levels of biomarkers for 10 μg/m³ increases in PM_2.5 and NO₂ and error bars represent 95% confidence intervals. The red lines and error bars refer to significant associations and the black lines and error bars refer to non-significant associations.

We further explored the potential effect modification by sex, smoking, BMI, and CAC score in the associations of NO₂ exposure with CRP, ApoA, TG, and glucose (Figure 4 and Figure S4). Association between NO₂ and CRP was stronger among females, non-smokers, and individuals without CAC as compared to males, smokers, and those with a CAC score >0. We observed stronger associations of NO₂ exposure with ApoA among those who were smokers, with BMI <25 kg/m², and no CAC. Associations of NO₂ exposure with TG and glucose were stronger among males. In addition, significant NO₂-glucose associations were found among non-smokers and participants with BMI <25 kg/m² but not among those who were smokers and with BMI ≥25 kg/m².

Figure 4. The single-day and cumulative effects of exposure to nitrogen dioxide (NO₂) on levels of biomarkers in stratified analyses.

ApoA, apolipoprotein A; hsCRP, high-sensitivity C-reactive protein; BMI, body mass index; CAC, coronary artery calcium. Models were adjusted for age (continuous), sex (male, female), education (<college, college, >college), BMI (continuous), smoking status (never/past, current), smoking years (continuous), cigarettes per day (continuous), alcohol use (no, yes), physical activity (never, ≤3 times/wk, >3 times/wk), urbanization (urban, rural), payment method (fully self-paid, partially self-paid, fully-covered), antihypertensive medication use (no, yes), statin use (no, yes), temperature (df=3), and humidity (df=3). Dots represent percent difference percent difference in levels of biomarkers for a 10 μg/m³ increase in NO₂ and error bars represent 95% confidence intervals. The red lines and error bars refer to significant associations and the black lines and error bars refer to non-significant associations. * indicates significant heterogeneity in the associations between subgroups.

In two-pollutant models, the results were generally similar to the primary analysis, although some estimates varied slightly to some extent (Figure S5 and Figure S6).

4. Discussion

In this cross-sectional study, we identified different patterns of multiple cardiometabolic biomarkers in response to variable lag days and durations of exposure to PM_2.5 and NO₂ in a susceptible population of patients suspected with CAD. In the single-day models, we found a strong but attenuated trend of associations of PM_2.5 and NO₂ with TC, ApoA, and ApoB from the first lag day of exposure, whereas for hs-CRP and HDL-C, the significant exposure windows presented after the first few lag days. In the cumulative models, air pollution exposure was associated with changes in ApoA over up to 30 week-average, hs-CRP over an 8-week average, and TG, glucose, TG/HDL-C, and TyG index over a one-week average. Our study suggested that short- and medium-term, rather than long-term, exposure to air pollution may be associated with changes in a variety of cardiometabolic biomarkers.

PM2.5 exposure can induce more severe inflammation and oxidative stress in the circulation system of hyperlipidemic rats, promote a hypercoagulable state, and trigger cardiomyocyte apoptosis. CRP is a marker of systemic inflammatory response and one of the predictors for CVD (Araujo, 2011; Brook et al., 2010; Rifai and Ridker, 2001). Air pollutants may trigger cellular signaling networks, activate immune cells, and further elevate CRP levels within the lungs and throughout the body (Glencross et al., 2020; Lakhdar et al., 2022; Xu et al., 2022b). In this study, hs-CRP levels increased significantly by up to 15.85% with increased NO₂ concentrations averaged over 8 weeks before the blood draw, but the association became attenuated to null for longer periods. These findings were supported by a recent meta-analysis, in which CRP was associated with short-term exposure to NO₂ over days or weeks rather than longer-term exposure averaged over 6 months (Xu et al., 2022b). We also observed short-term effects of PM_2.5 on hs-CRP levels, but the associations were attenuated when considering long-term exposures. This finding was slightly different from a meta-analysis by Liu et al., in which stronger associations for long-term (over 6 months) than short-term exposures to PM_2.5 on CRP were identified (Liu et al., 2019b). This might be because our study population was relatively unhealthy (patients with relatively high background hs-CRP levels) and therefore were more susceptible to acute exposure, especially in a highly-polluted area (Liu et al., 2019b). Moreover, averaging exposure over a longer period may smooth out the temporal variation of the exposure, resulting in a smaller exposure variation (e.g. SDs of PM_2.5 levels averaged over 1-wk and 1-year: 45.69 and 3.53 μg/m³, respectively) and greater uncertainty (i.e., wider CIs) in association with the cardiometabolic outcomes.

In the stratified analysis, stronger associations were observed among females and nonsmokers, which were consistent with previous studies (Dadvand et al., 2014; Iyer et al., 2022; Panasevich et al., 2009). Compared to nonsmokers, smokers might have an upregulation of anti-inflammatory response to environmental stimuli due to long-term chronic inflammation (Frampton et al., 1997) and nicotine has also been found to exert anti-inflammatory activities in vitro and in vivo (Gonçalves et al., 2011; Lakhan and Kirchgessner, 2011). The majority of female participants were nonsmokers in our study (95.7%), which could explain the stronger effects among females, as the associations became similar when further stratified by sex among nonsmokers only in our study. In addition, we found stronger effects among individuals without CAC. Several lines of evidence have suggested that inflammation plays a key role in the development of atherosclerosis and might be involved in coronary calcification (Bessueille and Magne, 2015; Li et al., 2007). Participants with a CAC score >0 in our study presented a higher level of CRP than individuals without CAC in our study. Individuals with subclinical atherosclerosis might be less sensitive regarding their inflammatory response to air pollution exposure.

PM_2.5 was found to disrupt lipid metabolism and induce atherosclerosis progression (Araujo, 2011; Boren et al., 2020; Yang et al., 2021), and abnormal lipid profiles have been an important indicator for risk of atherosclerosis and CVD (Kjeldsen et al., 2022; McGuinn et al., 2019). We observed negative associations of air pollution exposure averaged over the first 6 days before the blood draw with TC and TG, and this was supported by some previous studies (Chuang et al., 2010; Wu et al., 2020) but different from others reporting associations for medium- or long-term exposure (Yitshak Sade et al., 2016; Zhang et al., 2021). ApoA is a major component of HDL particles and a protective factor against CVD, due to its role in reverse cholesterol transport (Fielding et al., 2000). Our findings of significant inverse associations of PM_2.5 and NO₂ exposure averaged up to 30 weeks with ApoA levels were in line with previous studies reporting both short- and medium-term protective effects of ApoA against air pollution exposure (Li et al., 2019; Wu et al., 2019). Although we did not find significant associations for LDL-C levels in any time lags, short-term effects over the prior one week were found for ApoB, which is a structural component of LDL particles and other atherogenic lipoproteins (e.g., chylomicrons very low-density lipoproteins) and was found to be a more accurate predictor for cardiovascular risk than LDL-C alone (Marston et al., 2022; Pencina et al., 2020; Sniderman et al., 2011). This finding confirmed the short-term effects of PM and NO₂ on ApoB (Chuang et al., 2010; Xu et al., 2022a), while the effects of longer lag days need further investigation (Mao et al., 2020; McGuinn et al., 2019; Wu et al., 2019). In the cumulative effect models, NO₂ exposure over longer averaging periods was unexpectedly associated with higher HDL-C levels. Due to its protective properties against atherosclerosis, the increase in HDL might suggest metabolic adaptations due to long-term air pollution exposure (Yang et al., 2021). Although similar results were reported in previous studies in China and North Carolina (McGuinn et al., 2019; Xiao et al., 2016; Zhang et al., 2021), other epidemiologic studies supported a negative or null association (Bell et al., 2017; Wu et al., 2019; Wu et al., 2020; Yitshak Sade et al., 2016). Our findings might result from chance due to multiple comparisons and should be interpreted with caution until further investigation. Overall, the lack of significant findings of long-term effects on lipids/apolipoproteins might be because our study population was atherosclerotic patients, who may be more susceptible to acute harmful effects of air pollution, be more likely to seek health care following short-term exposure and therefore be included in our study. The significant results observed in the short term but not in the long term may also be due to larger variations of exposures averaged over shorter periods than that over longer duration. Further studies are warranted to provide more evidence on how lipid metabolism can be affected by air pollution over different duration of exposure in susceptible population.

Accumulating evidence has linked air pollution exposure with abnormal glucose metabolism (Chen et al., 2016b; Teichert et al., 2013; Yitshak Sade et al., 2016). In line with previous studies (Chen et al., 2016a; Peng et al., 2016; Zhan et al., 2021), we observed increased fasting glucose levels associated with acute exposure to higher air pollution levels (averaged over 6 days). However, increased glucose levels were also related to longer averaging periods in other studies (Lucht et al., 2018; Peng et al., 2016; Tian et al., 2022; Yitshak Sade et al., 2016). Lucht et al. reported that PM_2.5-glucose associations were significant and strongest during 28- to 60-day exposure windows before the blood measures among 7,108 nondiabetic participants in the Heinz Nixdorf Recall Study (Lucht et al., 2018). Inconsistent findings might result from differences in population characteristics, study designs, or different sources and compositions of ambient PM_2.5. Notably, serum glucose may not be an ideal outcome measure for assessing the potential long-term effects of air pollution exposures since it can vary widely over short periods. Biomarkers of long-term glycemic control, such as glycosylated hemoglobin, might be more robust to reflect the effects of air pollution exposure over the longer duration. TG/HDL-C ratio and TyG index have been proposed as reliable surrogate markers of insulin resistance (Giannini et al., 2011; Guerrero-Romero et al., 2010; Simental-Mendía et al., 2008). Similar to results from the Women’s Health Initiative (Holliday et al., 2019), we did not observe significant associations of 12-month average levels of PM_2.5 and NO₂ with TG/HDL-C ratio or TyG index. However, we identified positive associations for the one-week exposure window. To date, the relationships between air pollution exposure and TG/HDL-C ratio or TyG index are still understudied, and therefore more studies on these associations that explore the important exposure window are needed to provide insights into how air pollutants result in impaired glucose metabolism.

The major strength of our study is the inclusion of extensive cardiometabolic biomarkers and the flexible exposure time intervals applied in the susceptible population. Moreover, the use of high temporal (daily) and spatial (100m) resolution in estimates of PM_2.5 and NO₂ decreased the potential exposure misclassifications. Last, we were able to control for several important confounding variables, such as the use of medications and lifestyle factors (e.g., physical activity, smoking).

Several limitations have to be acknowledged. First, our study was cross-sectional with a relatively small sample size. The blood samples were taken at a single point at a time; therefore, a snapshot analysis may not provide valid inference on the lagged and cumulative effects of air pollution on cardiometabolic biomarkers. However, a single-point blood draw may only lead to random error with a possible bias towards the null. Future longitudinal studies with repeated measurements can further elucidate the trends in biomarkers following air pollution exposure. Second, this study was performed in patients with suspected CAD, and the results may not be generalizable to a healthier population.

5. Conclusions

The impacts of air pollution exposure on inflammatory markers, lipid profiles, and glucose levels varied by lag days and averaging periods of exposure, suggesting different underlying mechanisms through which air pollution may cause CVD. Our findings of the short- and medium-term, but not long-term effects of exposure to PM_2.5 and NO₂ on lipids/lipoproteins, CRP, and glucose levels merit more epidemiological studies to further confirm air pollution-induced inflammatory changes and alterations in lipid and glucose metabolism.