PM_2.5 and its chemical components interact with hypoxia to increase the risk of hospitalization for peripheral arterial disease

Abstract

This study investigated the combined effects of fine particulate matter (PM_2.5) exposure and ambient oxygen (O₂) availability on peripheral artery disease (PAD) hospitalizations in Gansu Province, China, from 2018 to 2022 (N = 38,514). A time-series case-crossover design with conditional logistic regression assessed short-term exposure to daily PM_2.5 and its major chemical components (black carbon [BC], organic carbon [OC], sulfate [SO₄²⁻], nitrate [NO₃⁻], ammonium [NH₄⁺], and chloride [Cl⁻]). Atmospheric O₂ levels were estimated based on altitude, temperature, and humidity. Short-term exposure to PM_2.5 was associated with an increased risk of PAD hospitalization, notably at lag 1 (odds ratio [OR] = 1.003; 95% confidence interval [CI]: 1.002–1.004). Carbonaceous pollutants exhibited delayed effects, with peak associations at lag 6 for BC (OR = 1.069; 95% CI: 1.035–1.105) and OC (OR = 1.028; 95% CI: 1.011–1.050), per 1 μg/m³ increase. Secondary inorganic aerosols showed acute effects at lag 1: Cl⁻ (OR = 1.107; 95% CI: 1.063–1.150), NH₄⁺ (OR = 1.040; 95% CI: 1.022–1.060), SO₄²⁻ (OR = 1.025; 95% CI: 1.015–1.036), and NO₃⁻ (OR = 1.030; 95% CI: 1.018–1.042), per 1 μg/m³ increase. Nonlinear exposure–response curves revealed a stronger PAD risk under lower O₂ conditions (<18%). The PM_2.5-related risk was amplified in high-altitude residents (>1500 m), older adults (>60 years), females, emergency admissions, and during the cold season. These findings suggest that ambient PM_2.5 and lower O₂ levels interact synergistically to elevate PAD hospitalization risk, emphasizing the urgent need for region-specific air quality controls and targeted health protection strategies.

Graphical abstract

Highlights

• •Short-term exposure to ambient PM_2.5 significantly elevates the risk of hospitalization for peripheral artery disease (PAD).
• •The impact of PM_2.5 on PAD hospitalization is exacerbated by environmental hypoxia.
• •Specific chemical components of PM_2.5 exhibit distinct temporal patterns of association with PAD.
• •Females, older adults, and people in cold seasons are more susceptible to adverse effects of PM_2.5 on PAD risk.

1. Introduction

Peripheral arterial disease (PAD) is a prevalent circulatory disorder characterized by the narrowing or occlusion of peripheral arteries, primarily due to atherosclerosis, leading to reduced or interrupted blood flow [1]. This condition manifests as leg pain, impaired mobility, and critical limb ischemia, and is associated with a heightened risk of major adverse cardiovascular events, such as myocardial infarction and stroke [2]. The Global Burden of Disease Study shows that in 2019, there were 113 million PAD patients among individuals aged ≥40 years, with the highest prevalence in the 80−84 age group (14.91%) [3]. Notably, 69.4% of PAD-related disability-adjusted life years are attributed to modifiable risk factors. With its high prevalence, significant complications, and the considerable economic strain it imposes on societies worldwide, PAD has emerged as a critical global public health concern.

Emerging evidence indicates that air pollution, particularly fine particulate matter (PM_2.5), plays a key role in the onset and progression of PAD [4,5]. PM_2.5, defined as particles ≤2.5 μm in diameter, can penetrate deep into the respiratory and circulatory systems, triggering systemic inflammation and oxidative stress [6,7]. According to the World Health Organization, ambient air pollution caused about 4.2 million premature deaths globally in 2019, with PM_2.5 as the primary contributor [8]. Its composition includes organic carbon (OC), black carbon (BC), and secondary inorganic aerosols (sulfate [SO₄²⁻], nitrate [NO₃⁻], ammonium [NH₄⁺], and chloride [Cl⁻]) [9], originating from diverse sources with varying health effects. Northwestern China faces severe PM_2.5 pollution from both natural (e.g., dust storms) and anthropogenic (e.g., fossil fuel combustion) sources [10]. Studies have shown that PM_2.5 could exacerbate atherosclerosis by inducing endothelial oxidative stress, promoting pro-inflammatory cytokine release, and impairing vascular function [11,12]. However, most research focuses on total PM_2.5 mass, with limited attention to individual component effects.

Ambient oxygen (O₂) availability, a critical determinant of physiological processes and energy metabolism, is influenced by key physical factors such as the interplay of altitude, temperature, and humidity [13]. With increasing altitude, decreasing atmospheric pressure reduces the partial pressure of O₂, thereby limiting O₂ delivery to tissues [14]. Hypoxia not only amplifies oxidative stress but also disrupts endothelial repair mechanisms and promotes systemic inflammation, key drivers of atherosclerosis progression [15,16]. Moreover, specific chemical components of PM_2.5 may interact with hypoxic conditions to exacerbate their biological effects. For example, OC, NO₃⁻, SO₄²⁻, and NH₄⁺ can contribute to vascular inflammation and reduced nitric oxide bioavailability [17], and these effects may be further aggravated in hypoxic settings. Despite these potential interactions, few studies have systematically investigated how PM_2.5 exposure and lower O₂ availability collectively influence the risk of PAD.

This knowledge gap is particularly pronounced in Northwestern China, where the confluence of high-altitude conditions and severe PM_2.5 pollution creates a unique environmental context that may disproportionately affect vascular health. To address this gap, we conducted a time-series case-crossover study analyzing regional hospitalization records from 2018 to 2022 to examine the interactive effects of short-term PM_2.5 exposure and its chemical constituents, as well as ambient O₂ levels on the risk of PAD hospitalization. By elucidating the complex interplay among these environmental factors, our findings aim to inform evidence-based public health interventions and improve health outcomes for vulnerable populations residing in high-altitude regions.

2. Methods

2.1. Study population

This study analyzed PAD-related hospital admissions from 375 hospitals in Gansu Province, China, from 2018 to 2022, using a provincial healthcare database. PAD cases were identified via the International Classification of Diseases, 10th Revision (ICD-10) codes G45.1, G45.2, I70–I79, and I65. These codes capture a wide range of PAD-related diagnoses, ensuring comprehensive case inclusion consistent with methodologies employed in prior studies utilizing administrative healthcare data [18,19]. Acute events were classified per the criteria detailed in Table S1; others were considered non-acute. Patient-level data included sex, age, admission/discharge dates, and admission route (emergency, outpatient, transfer, or other). Hospital coordinates were obtained via the Baidu Maps API to approximate patient residential locations, enabling environmental exposure assignment. Fig. S1A shows the spatial distribution of included hospitals and PAD admissions. This study protocol was approved by the Institutional Review Board of the School of Public Health, Lanzhou University (IRB24090503). The need for informed consent was waived since personal identification information was not provided.

2.2. Environmental data

To assess environmental exposure for each hospitalization event, we estimated daily concentrations of PM_2.5 and its major chemical components at the hospital level. Specifically, we obtained gridded daily air pollution data from three complementary datasets: NASA’s Modern-Era Retrospective Analysis for Research and Applications version 2 (MERRA-2), China High Air Pollutants (CHAP), and Tracking Air Pollution in China (TAP). Each dataset offers distinct spatial and temporal resolutions and pollutant coverage, enabling a multi-faceted analysis that enhances result reliability through multi-dataset validation and compensates for individual dataset limitations.

MERRA-2 Dataset: The MERRA-2 dataset (https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/) is a global atmospheric reanalysis product provided by NASA [20,21]. This dataset provides robust and continuous estimates of key aerosol components, including PM_2.5, OC, and BC, with hourly data available at a spatial resolution of 0.5° × 0.625°. Notably, MERRA-2 derives PM_2.5 through empirical formulas at ambient relative humidity with hygroscopic growth applied to sulfate and sea salt, whereas OC and BC are reported as dry mass without explicit size fractionation. Although not formally size-resolved to PM_2.5, these carbonaceous components are widely validated as reliable indicators of fine-fraction aerosol exposure in epidemiological studies [22]. Due to its broad spatial coverage and temporal continuity, MERRA-2 is particularly valuable in supplementing exposure data in rural or data-sparse regions, such as parts of western China, including Gansu Province.

CHAP Dataset: The CHAP dataset (https://weijing-rs.github.io/product.html) provides high-resolution (1 km × 1 km) daily estimates of PM_2.5 and secondary inorganic aerosols, specifically SO₄²⁻, NO₃⁻, NH₄⁺, and Cl⁻, as well as other pollutants such as ozone (O₃). These chemical species are present in the PM_2.5 fraction and are modeled using an advanced data fusion algorithm integrating ground-based monitoring, satellite remote sensing, and chemical transport modeling [[23], [24], [25], [26]]. CHAP offers the finest spatial resolution among the three datasets, making it particularly suitable for high-precision exposure estimation in urban and peri-urban areas. However, its coverage is geographically limited to certain regions (e.g., selected cities in Gansu Province), necessitating supplementation from larger-scale datasets.

TAP Dataset: The TAP dataset (http://tapdata.org.cn/) was developed by Tsinghua University through the integration of ground monitoring data, satellite retrievals, emission inventories, and chemical transport model simulations [27]. It provides daily concentrations of PM_2.5 and associated chemical components, including both primary particulates, such as organic matter (OM) and BC, and secondary inorganic aerosols (SO₄²⁻, NO₃⁻, and NH₄⁺), at a resolution of 10 km × 10 km [28]. This comprehensive composition allows for a robust analysis across diverse geographical areas.

Furthermore, daily meteorological variables, including temperature and relative humidity (RH), were derived from the ERA5 reanalysis dataset [29]. This dataset was used to control for the confounding effects of meteorological factors in the statistical analysis. This multi-dataset validation study strategy effectively utilizes the high spatial resolution of CHAP, the detailed pollutant composition of TAP, and the broad spatial and temporal coverage of MERRA-2 to ensure robust and reliable estimates of the effects of environmental exposures while mitigating the shortcomings of any individual dataset.

For the exposure assignment, the geographic coordinates (latitude and longitude) of each hospital were first determined. For each dataset, we implemented a spatial matching algorithm to link each hospital to the nearest grid cell centroid. This procedure aimed to assign the most representative environmental data. Subsequently, daily mean values for each pollutant and meteorological variable were calculated for every grid cell corresponding to a hospital. These daily time-series data were then assigned as the exposure values for all hospitalization events recorded at that specific hospital on a given day. This method provided a comprehensive, daily-resolved exposure profile for each hospital throughout the study period.

2.3. Atmospheric O₂ concentration estimates

To quantify effective atmospheric O₂ availability in Northwest China, we calculated altitude-adjusted equivalent O₂ concentrations accounting for altitude, temperature, and relative humidity (RH), factors that influence inspired O₂ partial pressure [13]. Altitude for each hospital’s location was determined by spatially matching its geographic coordinates (latitude and longitude) with data from the AWS Terrain Tiles dataset [30]. Fig. S1B shows the spatial distribution of included hospitals and altitude. The corresponding actual O₂ concentration (O₂, %) for each patient was calculated using the following equation:

$$O_2=\left(\frac{P_{\text{actual}}-P_{{\mathrm{H}}_2\mathrm{O}}^{\text{actual}}}{1013.25}\right)\times 0.20946\times \left(\frac{293.15}{273.15+T}\right)\times 100\%$$ (1)

$$P_{\text{actual}}=1013.25\times e^{-h/7400}$$ (2)

$$P_{{\mathrm{H}}_2\mathrm{O}}^{\text{actual}}=P_{{\mathrm{H}}_2\mathrm{O}}\times RH$$ (3)

$$P_{{\mathrm{H}}_2\mathrm{O}}=6.1078\times {10}^{\left(\frac{7.5T}{237.3+T}\right)}$$ (4)

Where, P_actual denotes actual atmospheric pressure (hPa), derived from the barometric formula [13], h denotes altitude above sea level (m); $P_{{\mathrm{H}}_2\mathrm{O}}^{\text{actual}}$ denotes actual water vapor pressure (hPa); $P_{{\mathrm{H}}_2\mathrm{O}}$ denotes saturated water vapor pressure (hPa), estimated using the Magnus-Tetens formula [31]; RH and T denote relative humidity (decimal form) and ambient temperature (°C), respectively; 0.20946 is the volumetric fraction of O₂ in dry air; 293.15/(273.15 + T) denotes temperature correction factor, normalizing O₂ concentration to a standard temperature of 20 °C.

2.4. Statistical analysis

To assess the association between short-term exposure to high-resolution PM_2.5, its chemical constituents, and atmospheric O₂ concentrations on the risk of hospitalization for PAD in Northwest China, we employed a time-series matched case-crossover study design [32]. In this design, each “case day” was defined as the date of a patient’s PAD hospitalization. This design prioritizes sensitivity to short-term exposure variations and is commonly used in air pollution studies of acute health events, especially when the exposure shows high day-to-day variability. Environmental exposure levels on the case day were compared with those on control days, selected as 2–4 days before the case day, with each case day typically matched to three control days. This self-matching approach controls for individual-level, time-invariant confounders (e.g., sex, age, and lifestyle), reducing bias from demographic and short-term risk factors.

We used conditional logistic regression models to estimate associations between environmental exposures (PM_2.5, its chemical constituents, and atmospheric O₂ concentrations) and PAD hospitalization risk. The model adjusted for time-varying covariates, including daily temperature, RH, and O₃. Nonlinear effects of temperature and RH were modeled using natural cubic splines with 3 degrees of freedom (df), determined by the minimum Bayesian information criterion (BIC). The conditional logistic regression model is expressed as:

$$\begin{array}{l}\text{logit}\left[P\left(Y_{it}=1\right)\right]={\beta }_0+{\beta }_1{PM}_{2.5,it}+\text{ns}\left(T_{it},\text{df}=3\right)+\text{ns}\left({HR}_{it},\text{df}=3\right)\\ +{\beta }_2{O}_{3,it}+\text{strata}\left(i\right)\end{array}$$ (5)

Where, P(Y_it = 1) represents the probability that hospitalization of the i-th case occurs at time t; β₀ represents intercept; β₁ and β₂ represent regression coefficients for PM_2.5, its chemical constituents and O₃, respectively; PM_2.5,it and O_3,it represent the concentrations of PM_2.5 (μg/m³) and O₃ (μg/m³) for the i-th case at time t, respectively; ns(T_it, df = 3) represents natural spline for temperature with 3 df; ns(HR_it, df = 3) represents natural spline for RH with 3 df; strata (i) represents the stratum for the i-th case, used to control for time-invariant individual characteristics through the case-crossover design.

We estimated the odds ratios (ORs) and 95% confidence intervals (CIs) to evaluate the association between pollutant exposure and PAD hospitalization. Effect estimates for single-day lags of PM_2.5 and its components were scaled to a 1 μg/m³ increase, while in the subgroup analyses, the estimates for cumulative PM_2.5 exposure were scaled to a 10 μg/m³ increase. To examine lagged and cumulative effects, we refitted the model with single-day lags (lag 0 to lag 7) and multi-day moving averages (e.g., lag 0–1 for current and previous day exposures), selecting the optimal lag window based on model fit.

We examined potential non-linear dose-response relationships between ambient concentrations of PM_2.5, its chemical constituents, atmospheric O₂, and the risk of PAD hospitalization using restricted cubic splines (RCS). The optimal number of knots for each model was selected by minimizing the BIC. For O₂, the RCS model utilized three knots placed at the 10th, 50th, and 90th percentiles of its distribution. For subsequent stratified and interaction analyses, O₂ concentrations were dichotomized into hypoxia (<18%) and normal (≥18%), with the threshold for hypoxia being defined according to the European Spallation Source guideline for oxygen deficiency hazard [33].

Given the observed non-linear associations between PM_2.5, its constituents, ambient O₂, and the risk of PAD hospitalization, we employed three-dimensional response surface plots to characterize their joint effects and capture potential interactions across the exposure range. To isolate the combined impact of PM_2.5 and O₂, other meteorological covariates (RH and O₃) were fixed at their mean values over the study period. Linear predictors (log-odds) were estimated for each combination of PM_2.5 and O₂ using the fitted model. A reference scenario, defined by the median values of PM_2.5 and O₂, was used to calculate baseline log-odds. Relative risks were derived as ORs using the expression:

$$OR\left(x_1{,x}_2\right)=\exp \left[\eta \left(x_1{,x}_2\right)-\eta \left(x_1^{ref},x_2^{ref}\right)\right]$$ (6)

Where, $x_1$ and $x_2$ denote concentrations of PM_2.5 (μg/m³) and O₂ (%), respectively, and $\eta$ (⋅) is the model-predicted linear predictor.

The resulting OR matrix was visualized via interactive three-dimensional response surface plots, facilitating the identification of synergistic effects and potential effect modification by ambient O₂, thereby providing a comprehensive view of PM_2.5-related PAD risk across varying O₂ levels.

We further explored the heterogeneity of the association between air pollution and PAD hospitalization by conducting stratified analyses to identify potential effect modifiers. The analyses were stratified by age (<60 vs. ≥60 years), sex (male vs. female), season (warm [April–September] vs. cold [October–March]), admission route (emergency vs. outpatient), altitude (>1500 m vs. ≤1500 m), and PAD clinical presentation (acute vs. non-acute). The statistical significance of differences in effect estimates between subgroups was formally assessed using a Z-test. The Z-statistic was calculated by comparing the effect estimates (Coef) and their standard errors (SE) from the two respective subgroups:

$$Z=\frac{{Coef}_1-{Coef}_2}{\sqrt{{SE}_1^2+{SE}_2^2}}$$ (7)

Where, Coef₁ and Coef₂ are the estimated effects for the two subgroups, and SE₁ and SE₂ are their respective standard errors. The resulting Z was compared against a standard normal distribution [N(0, 1)] to obtain a two-sided P value. A P value < 0.05 was considered to indicate a statistically significant difference in the association between the two groups.

We conducted sensitivity analyses to address multicollinearity among PM_2.5 components. First, for component groups with high intercorrelation (Spearman r > 0.80), we applied principal component analysis (PCA) to derive orthogonal composite variables (Table S2). For components with moderate correlations (r < 0.80), we constructed multi-pollutant models including all components simultaneously. Second, to evaluate whether component effects were independent of total PM_2.5 mass, we employed two-pollutant models adjusting for PM_2.5. For components highly correlated with PM_2.5 (r > 0.80), we additionally conducted residual-based analyses; for those with lower correlations, standard two-pollutant models were used. All analyses were based on conditional logistic regression models, adjusted for daily mean temperature, RH, and O₃.

All statistical analyses were conducted in R software (version 4.3.1), with two-sided tests and a significance threshold of P < 0.05.

3. Results

3.1. Characteristics of the study population and exposure distribution

A total of 38,514 hospitalized PAD patients were included in the study, with 62.1% male (n = 23,931) and 37.9% female (n = 14,583) (Table 1). Over half (53.0%, n = 20,408) were aged >60 years. Most hospitalizations (62.6%, n = 24,092) occurred at high-altitude locations (>1500 m), and slightly more admissions occurred during the warm season (53.8%). Non-acute cases accounted for 70.5%, and 58.9% were admitted via outpatient visits. The majority of hospitalizations (74.6%, n = 28,727) occurred at locations where atmospheric O₂ levels were below 18%. Average air pollutant and meteorological exposures were similar on case and control days (Table S3–S5).

Table 1.

Characteristics of peripheral arterial disease cases in this study.

Population characteristics	Number	Proportion
Total	38,514	100.0%
Sex
Male	23,931	62.1%
Female	14,583	37.9%
Age
≤60 years	18,106	47.0%
>60 years	20,408	53.0%
O2
<18%	28,727	74.6%
≥18%	9787	25.4%
Altitude
≤1500 m	14,422	37.4%
>1500 m	24,092	62.6%
Year
2018	6272	16.3%
2019	8504	22.1%
2020	7360	19.1%
2022	8198	21.3%
2021	8180	21.2%
Season
Warm	20,721	53.8%
Cold	17,793	46.2%
PAD clinical presentation
Acute exacerbation	11,367	29.5%
Non-acute	27,147	70.5%
Admission route
Emergency	11,990	31.1%
Outpatient	22,703	58.9%
Transfer	212	0.6%
Other	3609	9.4%

3.2. Associations between PM_2.5, its chemical components, and risk of hospitalization for PAD

Fig. 1 shows associations between short-term exposure to MERRA-2 pollutants and PAD hospitalizations. Exposure-response curves revealed dose-dependent increases in risk with higher PM_2.5, BC, and OC levels (Fig. 1A–C). The single-day lag analysis further revealed distinct temporal patterns (Fig. 1D–F). For each 1 μg/m³ increase in PM_2.5, a significant and immediate increase in risk was observed from lag 1 to lag 4 (OR_lag4 = 1.002; 95% CI: 1.001–1.003). In contrast, the effects of its carbonaceous components were delayed. For each 1 μg/m³ increase, the risks associated with BC and OC became significant only at later lags, with the strongest association observed for BC at lag 6 (OR = 1.069; 95% CI: 1.035–1.105) and for OC also at lag 6 (OR = 1.028; 95% CI: 1.011–1.050).

Fig. 1.

Dose-response relationships and lagged effects of PM_2.5, BC, and OC from the MERRA-2 dataset on hospitalization for peripheral artery disease. Panels (A)−(C) illustrate the overall cumulative exposure−response relationships for PM_2.5, BC, and OC, respectively. The solid lines represent the odds ratios (ORs) and the shaded areas represent the 95% confidence intervals (CIs). The background histograms show the frequency distribution of daily pollutant concentrations. Panels (D)−(F) show the estimated OR (95% CI) from single-day lag models for a 1 μg/m³ increase in PM_2.5, BC, and OC, respectively. The points represent the OR estimates, and the vertical lines indicate the 95% CIs for each lag day from admission (lag 0) to 7 days prior (lag 7). All models were adjusted for daily mean temperature, relative humidity, and ozone.

Exposure-response curves revealed positive dose-dependent relationships for PM_2.5 and all four components (Fig. 2). Single-day lag analyses further demonstrated that these risks manifested rapidly, with all components showing statistically significant associations at lag 1 (Fig. 3). Specifically, each 1 μg/m³ increase in total PM_2.5 was associated with a 0.3% increase in hospitalization risk at lag 1 (OR = 1.003; 95% CI: 1.002–1.004), and this elevated risk persisted over multiple subsequent days. Similar immediate effects were observed for individual secondary inorganic components: Cl⁻ (OR = 1.107; 95% CI: 1.063–1.150), NH₄⁺ (OR = 1.040; 95% CI: 1.022–1.060), SO₄²⁻ (OR = 1.025; 95% CI: 1.015–1.036), and NO₃⁻ (OR = 1.030; 95% CI: 1.018–1.042).

Fig. 2.

Dose−response relationships between PM_2.5, secondary inorganic aerosols from the CHAP dataset, and the risk of peripheral artery disease hospitalization. The plots illustrate the overall cumulative exposure−response associations for (A) PM_2.5, (B) SO₄²⁻, (C) NO₃⁻, and (D) NH₄⁺, and (E) Cl⁻. The solid lines represent the estimated ORs and the shaded areas represent the corresponding 95% CIs. All models were adjusted for daily mean temperature, relative humidity, and ozone.

Fig. 3.

Single-day lag effects of PM_2.5 and secondary inorganic aerosols from the CHAP dataset on peripheral artery disease hospitalization. The plots show the estimated ORs and 95% CIs for PAD hospitalization associated with a 1 μg/m³ increase in the concentration of (A) PM_2.5, (B) SO₄²⁻, (C) NO₃⁻, and (D) NH₄⁺, and (E) Cl⁻. The points represent the OR estimates, and the vertical lines indicate the 95% CIs. The effects are shown for exposures on the day of admission (lag 0) up to seven days prior (lag 7). All estimates were derived from conditional logistic regression models, adjusted for daily mean temperature, relative humidity, and ozone.

To validate and extend findings from the MERRA-2 and CHAP datasets, we conducted additional analyses using the TAP dataset. Consistent with primary results, exposure-response analyses confirmed positive, dose-dependent associations between PAD hospitalization risk and short-term exposure to PM_2.5, BC, OM, and secondary inorganic aerosols (Fig. S2). Lag-response patterns observed in TAP largely aligned with MERRA-2 and CHAP (Fig. S3). OM showed delayed effects, consistent with OC in MERRA-2, while BC exhibited both immediate and delayed associations, reinforcing its biphasic response. Among inorganic components, SO₄²⁻ showed delayed risk elevations, whereas NO₃⁻ and NH₄⁺ demonstrated both immediate and delayed effects, similar to CHAP findings.

3.3. Interaction of PM_2.5, its chemical components and O₂ on the risk of PAD hospitalization

To assess the modifying effect of atmospheric O₂ on PM_2.5–PAD associations, we conducted stratified analyses using an 18% O₂ threshold to define hypoxia. As shown in Fig. 4A, PAD hospitalization risk increased nonlinearly with decreasing O₂ levels. Fig. 4B presents stratified effect estimates for cumulative exposure: per 10 μg/m³ increase in PM_2.5 and each 1 μg/m³ increase in its components. Across both CHAP and MERRA-2 datasets, associations were stronger under hypoxia (O₂ < 18%). In CHAP, the OR for PM_2.5 was 1.11 (95% CI: 1.07–1.15) under hypoxia, vs. 1.00 (95% CI: 0.94–1.06) under normal O₂ (P < 0.001). Significant effect modification was also observed for SO₄²⁻ (hypoxia: OR = 1.14, 95% CI: 1.10–1.18; normal: OR = 1.02, 95% CI: 0.99–1.05; P = 0.002) and NO₃⁻ (hypoxia: OR = 1.18, 95% CI: 1.13–1.24; normal: OR = 1.01, 95% CI: 0.99–1.04; P < 0.001). The strongest effect modifications were seen for NH₄⁺ (hypoxia: OR = 1.38, 95% CI: 1.28–1.50; P < 0.001) and Cl⁻ (hypoxia: OR = 1.52, 95% CI: 1.33–1.74; P < 0.001). In contrast, BC and OC showed no significant effect modification by O₂ levels. To further explore interactive effects, we constructed three-dimensional response surfaces (CHAP: Fig. 5; MERRA-2: Fig. S4; TAP: Fig. S5). In all datasets, PAD hospitalization risk increased with both higher pollutant levels and lower O₂, highlighting reduced ambient O₂ availability as a key effect modifier.

Fig. 4.

Associations between atmospheric oxygen concentration and peripheral artery disease (PAD) hospitalization risk, and stratified effects of air pollutants under hypoxic vs. normal oxygen conditions. (A) The exposure−response relationship between atmospheric O₂ concentration and the risk of PAD hospitalization. The solid red line indicates the OR with its 95% CI across the range of O₂ concentrations, and the shaded area represents the 95% CI. The vertical dashed line at 18% O₂ marks the threshold used to define hypoxia. The green density curve (right axis) shows the distribution of daily atmospheric O₂ concentrations. (B) Stratified analyses of the short-term cumulative effects of PM_2.5 and its major chemical components on PAD hospitalization, conducted under hypoxic (O₂ < 18%) and normal (O₂ ≥ 18%) conditions separately, using data from the MERRA-2 and CHAP datasets. The effect of PM_2.5 reflects per 10 μg/m³ increase, whereas the effects for chemical components reflect per 1 μg/m³ increase. P-values represent tests for statistical interaction between subgroups, calculated using the Z-test for differences in effect estimates. All estimates were derived from conditional logistic regression models, adjusted for daily mean temperature, relative humidity, and ozone.

Fig. 5.

Three-dimensional response surfaces showing the joint associations of short-term air pollutant exposure and ambient oxygen concentration with the risk of peripheral artery disease hospitalization, based on the CHAP dataset. Response surfaces depict estimated ORs relative to the median pollutant level, across varying levels of atmospheric O₂ concentration (%) and each pollutant: (A) PM_2.5, (B) SO₄²⁻, (C) NO₃⁻, (D) NH₄⁺, and (E) Cl⁻. ORs were derived from conditional logistic regression models adjusted for daily mean relative humidity and ozone. Predicted values are scaled to a reference point defined by the median pollutant level and median O₂ concentration, with other covariates held at their mean. Models incorporated cumulative exposure estimates, and surfaces were generated using a prediction grid.

3.4. Interaction of PM_2.5, its chemical components, and altitude on the risk of PAD hospitalization

To investigate altitude as a potential effect modifier, we performed stratified exposure–response analyses and heterogeneity tests. In the MERRA-2 dataset (Fig. S6), stronger associations between air pollutants and PAD hospitalization were observed at high altitude (>1500 m) for PM_2.5, BC, and OC. Similar patterns appeared in CHAP (Fig. S7), where SO₄²⁻, NO₃⁻, and NH₄⁺ also showed steeper exposure–response curves at higher altitudes. The ORs for BC and OC were 1.31 (95% CI: 1.21–1.42) and 1.18 (95% CI: 1.12–1.24), respectively, versus 1.11 (95% CI: 1.04–1.18) and 1.04 (95% CI: 1.00–1.09) at lower altitudes (Fig. S8).

3.5. Stratified analysis

Associations between PM_2.5 and PAD hospitalization were generally consistent across subgroups. Effect estimates were modestly higher among females (Fig. S9), older individuals (>60 years, Fig. S10), and during the cold season (Fig. S11), with statistically significant modification observed for several components, including SO₄²⁻, NO₃⁻, and NH₄⁺. In seasonal analyses, the effect of PM_2.5 in CHAP was significantly stronger in the cold season. In Fig. S12, emergency admissions showed a higher OR for PM_2.5 and all components compared to outpatient cases. In Fig. S13, no substantial effect modification was observed by disease classification.

3.6. Sensitivity analyses

The results of both sensitivity analyses were consistent with our main findings (Table S6). In the multi-component model using the PCA-derived factor representing NO₃⁻, NH₄⁺, and Cl⁻, the estimated associations remained positive and statistically significant. Similarly, in the models assessing the independent effects of individual PM_2.5 components after adjusting for PM_2.5, the findings consistently demonstrate significant effects.

4. Discussion

In this study, we systematically evaluated the short-term effects of ambient PM_2.5 and its chemical constituents on hospitalizations for PAD, with particular attention to the modifying role of atmospheric O₂ levels. Our results consistently demonstrated that short-term increases in PM_2.5 levels were significantly associated with heightened PAD hospitalization risk across multiple datasets. Both secondary inorganic aerosols (SO₄²⁻, NO₃⁻, NH₄⁺, and Cl⁻) and carbonaceous components (BC and OC) contributed to this association, although their temporal patterns differed. Notably, we found that lower atmospheric O₂ levels significantly amplified the association between PM_2.5 exposure and PAD hospitalization risk.

Our findings align with prior studies linking PM_2.5 exposure to increased cardiovascular risks, such as higher hospitalization rates for ischemic heart disease, stroke, and overall cardiovascular disease [34,35]. Our study further extends this association to PAD, enriching the body of evidence on the effects of PM_2.5 on the cardiovascular system. To evaluate the representativeness of our single-province study, we systematically compared pollutant concentrations in our study area with national and regional averages. The mean PM_2.5 concentration in our study area (33.58 μg/m³) was notably lower than the national average (46.3 μg/m³). Similarly, concentrations of major secondary inorganic aerosols in Gansu were consistently below national averages: Cl⁻ (1.52 vs. 2.0 μg/m³), SO₄²⁻ (6.73 vs. 9.8 μg/m³), NO₃⁻ (5.18 vs. 8.1 μg/m³), and NH₄⁺ (3.66 vs. 6.1 μg/m³) [35]. Despite these lower pollution levels, our case-crossover study found PM_2.5 exposure was significantly associated with an increased risk of PAD hospitalization at lag 1 (OR = 1.003 per 1 μg/m³ increase; 95% CI: 1.002−1.004). In Madrid, Spain, the relative risk (RR) of daily circulatory admissions among the elderly rose by 6.2% (RR = 1.062, 95% CI: 1.036–1.089) per 10 μg/m³ of PM_2.5 [36]. In contrast, other time-series studies have reported effects on daily admissions per 10 μg/m³ increase in PM_2.5. For instance, a study conducted in Beijing found a 0.30% increase in daily cardiovascular admissions per 10 μg/m³ increase in PM_2.5 [37], and in Wuhan, China, PM_2.5 exposure (lag 0–2) was linked to a 1.23% increase in daily cardiovascular admissions (95% CI: 1.01%–1.45%) [38]. Similarly, a study from the northeastern U.S. reported a 0.26% (95% CI: 0.08%–0.45%) increase in PAD admissions per 10 μg/m³ increase in PM_2.5 [39]. Notably, among all components, Cl⁻ exhibited the strongest association with PAD hospitalization risk, with a 10.7% increase per 1 μg/m³ at lag 1 (OR = 1.107, 95% CI: 1.063–1.150), contrasting prior evidence where it typically had weaker cardiovascular associations [35]. The stronger exposure-response relationship observed for Cl⁻ in Gansu may be attributed to several factors. First, Cl⁻ typically originates from region-specific emission sources such as waste incineration, coal combustion, and biomass burning [40]. Second, the average Cl⁻ concentration (1.52 μg/m³) exceeded that of BC (1.01 μg/m³), potentially contributing to its more pronounced health effects. Third, the higher spatial resolution of CHAP data used for Cl⁻ assessment, compared to the coarser MERRA-2 data used for BC exposure estimates, may have enabled more accurate exposure characterization. These findings underscore the importance of incorporating regional pollutant profiles into health risk assessments, particularly in areas with relatively low total particle mass where specific components may still pose significant health risks.

The direct deposition of PM_2.5-bound ions in the respiratory tract may trigger acute inflammatory and vascular responses relevant to PAD risk. The fine size and large surface area of PM_2.5 facilitate deep lung penetration and systemic translocation, triggering oxidative stress, vascular inflammation, and endothelial dysfunction—key processes in atherosclerosis and PAD development [41,42]. Beyond total PM_2.5, its chemical composition plays a critical role in toxicity. For example, BC, a core component of PM_2.5, induces oxidative stress and chronic inflammation, leading to endothelial dysfunction and accelerated atherosclerosis [43,44]. OC, often containing polycyclic aromatic hydrocarbons, may impair vascular signaling and promote smooth muscle migration [45,46]. Similarly, SO₄²⁻ is linked to arterial wall thickening and cardiovascular mortality, potentially through inflammatory and vascular pathways [47]. NO₃⁻ may disrupt nitric oxide signaling, contributing to vasoconstriction and impaired vascular function [48]. Studies have shown that NH₄⁺ and Cl⁻ may promote the release of inflammatory factors, exacerbating the inflammatory response in the body and causing vascular damage [49].

This study reveals that atmospheric O₂ and altitude are powerful effect modifiers that potentiate the adverse effects of PM_2.5 on the risk of hospitalization for PAD. Atmospheric O₂ concentration showed a significant, nonlinear inverse association with PAD hospitalization risk, suggesting that populations residing at high altitudes and experiencing chronic hypoxia are at persistently elevated risk. While adequate O₂ is essential for vascular homeostasis [50], with hypoxia affecting endothelial cell physiology and vasoactive substance regulation [51,52], our results indicate a synergistic interaction beyond hypoxia’s independent effects. Hypoxia may amplify the adverse vascular effects of PM_2.5 through several biological mechanisms. First, hypoxic conditions can upregulate hypoxia-inducible factors, which in turn modulate the expression of pro-inflammatory cytokines, adhesion molecules, and angiogenic factors, thereby promoting endothelial dysfunction and vascular inflammation [53,54]. Second, hypoxia impairs mitochondrial function and enhances reactive oxygen species production, which, when combined with PM_2.5-induced oxidative stress, may lead to synergistic endothelial damage and lipid peroxidation [55]. Third, hypoxia compromises antioxidant defense mechanisms and immune surveillance [56,57], leading to increased susceptibility to particle-induced inflammation and thrombosis. This is particularly relevant for individuals with pre-existing conditions such as chronic lung disease or hypoxic respiratory failure (e.g., severe COPD), who are more vulnerable to environmental pollutants [[58], [59], [60]]. Thus, the significant increase in the effect of PM_2.5 and its components on the risk of PAD hospitalization under hypoxic conditions is a concerning finding. The interaction between hypoxia and PM_2.5 may represent a pathophysiological synergy, where hypoxia sensitizes vascular tissues to PM_2.5-induced injury through enhanced oxidative stress, inflammation, and endothelial dysfunction.

Our stratified analyses revealed that associations between PM_2.5 exposure and PAD hospitalization risk were significantly stronger in specific subgroups. The heightened susceptibility observed in older adults (>60 years) is consistent with existing literature; age-related declines in vascular elasticity and immune function, often compounded by chronic comorbidities like hypertension and diabetes, can amplify sensitivity to pollution-induced oxidative stress and inflammation [61,62]. Interestingly, our data indicated a slightly higher risk for females. While some studies point to higher exposure for males due to occupational or commuting patterns [63], the stronger effect in females could be driven by hormonal factors, particularly the loss of estrogen’s vasculoprotective effects in the post-menopausal population, which aligns with the higher risk seen in older age groups [64]. The stronger associations during cold seasons likely reflect synergistic effects between low temperature and PM_2.5 exposure. Low temperatures are known to stimulate sympathetic activation and vasoconstriction, which can worsen peripheral artery ischemia [65]. When combined with PM_2.5 exposure, which induces endothelial dysfunction, the body’s ability to maintain adequate collateral circulation may be severely compromised. Notably, the stronger correlation found between emergency admissions and air pollution levels, as opposed to outpatient visits, offers a critical clinical insight into the immediate health impacts of environmental factors. This suggests that short-term air pollution exposure may act as an acute trigger for severe PAD events (e.g., acute limb ischemia) that require urgent care, rather than simply influencing scheduled appointments for chronic management. This finding underscores the immediate and clinically significant impact of air pollution on the most severe manifestations of PAD. Collectively, these stratified findings are critical for public health, as they help identify high-risk populations (older females) and high-risk conditions (cold weather, emergency presentations), enabling the development of targeted health advisories and clinical prevention strategies.

Our findings have important implications for populations in high-altitude regions. While our results suggest that higher ambient O₂ levels are associated with lower PAD risk, directly modifying outdoor atmospheric O₂ concentrations is not feasible. Therefore, the most practical and evidence-based strategy is to minimize PM_2.5 exposure in these vulnerable populations. This can be achieved through interventions such as indoor air purification, the use of protective masks during pollution episodes, and limiting outdoor activities on high-pollution days based on public health advisories. Regarding potential approaches to improve individual oxygenation status, clinical interventions such as supplemental oxygen therapy may theoretically help mitigate pollution-related risks in high-altitude residents with PAD, though this hypothesis requires validation through future controlled studies. Overall, our findings underscore the need for enhanced air quality management and targeted prevention strategies in hypoxic regions where pollution-related health impacts may be amplified.

This study has several limitations. First, due to its observational design, causal relationships between air pollutants and PAD hospitalizations cannot be established. Second, exposure misclassification may have occurred, as we relied on regional-level ambient air pollution data. Although we used multiple validated datasets (CHAP, TAP, and MERRA-2) to improve spatial and temporal accuracy, individual-level variations in exposure remain unmeasured. Future studies should consider personal exposure monitoring or high-resolution models. Third, the absence of individual-level sociodemographic and behavioral data may introduce residual confounding. While the case-crossover design helps control for time-invariant factors, unmeasured time-varying confounders could still affect results. Fourth, the study focused only on Gansu Province, which limits its generalizability; the applicability of these findings to regions with different environmental or demographic characteristics may require further investigation. Fifth, while atmospheric O₂ estimates were based on key environmental parameters (altitude, temperature, humidity), local factors like vegetation and traffic were not considered and may influence urban air quality [66,67]. Finally, our study focused on short-term effects. Long-term and cumulative impacts of O₂ and PM_2.5 interactions on PAD require further investigation through longitudinal research.

5. Conclusion

In summary, this study reveals that short-term PM_2.5 exposure significantly increases the risk of PAD hospitalization. The chemical composition of PM_2.5 is associated with this risk, with greater sensitivity in women, the elderly, and during colder months. Notably, the risk of PAD hospitalization is significantly amplified under conditions of reduced atmospheric O₂ availability, such as at high altitudes. These findings underscore the urgent need for targeted public health strategies and air quality regulations that account for not only pollutant mass but also specific chemical compositions and high-risk environmental conditions to mitigate the acute vascular impacts of air pollution.

CRediT authorship contribution statement

Li He: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis. Ce Liu: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis. Baofeng Zhou: Methodology, Data curation. Hao Zhao: Data curation. Zhaoru Yang: Data curation. Erkai Zhou: Data curation. Huan Chen: Data curation. Huanhuan Wei: Data curation. Ququmo Guoji: Resources, Project administration. Yuxin He: Resources, Project administration. Bin Luo: Writing – review & editing, Validation, Supervision, Project administration.

Availability of data and materials

Data will be made available on request.

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. Supplementary data

The following is the supplementary data to this article: