Association of joint exposure to PM2.5 and extreme heat with cardiovascular disease: A population-based cohort study in Northwest China

Summary

Extreme heat and fine particulate matter (PM_2.5) may increase cardiovascular disease (CVD) mortality, but their combined impact on overall CVD incidence is unclear. We aimed to evaluate their independent effects, identify vulnerable populations, and explore interactions, using a 2019 cohort of 5,070,689 adults. This year’s average PM_2.5 and extreme heat (annual ratio of extreme heat days (AREHDs), annual ratio of extreme heat hours (AREHHs)) exposures were analyzed via accelerated failure time (AFT) models. Each 1 μg/m³ PM_2.5 increase shortened CVD onset time by 0.083% (95% confidence interval [CI]: −0.009–0.175), most affecting the elderly, males, smokers, drinkers, and those with hypertension or diabetes. Each 1% of AREHD and AREHH increase reduced onset time by 0.446% (95% CI: 0.304–0.587) and 1.379% (95% CI: 1.043–1.714), stronger in the elderly, males, and sedentary individuals. PM_2.5 and extreme heat exerted synergistic detrimental effects, with the greatest risk observed at medium levels of co-exposure.

Graphical abstract

Highlights

• •A 5.07 million-participant cohort shows long-term PM_2.5 and extreme heat raise CVD risk
• •Male, sedentary individuals, smokers, and drinkers are prone to both PM_2.5 and extreme heat
• •PM_2.5 and extreme heat have a nonlinear synergistic effect on the onset time of CVD
• •The peak of the combined risk is located at moderate exposure levels

Health sciences

Introduction

Despite advancements in treatment and prevention, cardiovascular disease (CVD) remains the leading cause of global mortality.¹ Influenced by high-risk factors such as smoking, alcohol consumption, obesity, and diabetes, approximately four million individuals succumb to CVD annually in China, underscoring the intricate interplay among genetic susceptibility, environmental influences, and lifestyle factors.^2,3 This alarming statistic highlights the severe mortality rates and emphasizes the multifaceted impact of CVD on health, economic stability, and quality of life.^2,4 Furthermore, climate change exacerbates the risk of CVD through its contribution to extreme temperature fluctuations and declining air quality.⁵ Developing effective medical interventions and public health policies is crucial to address the urgent threat posed by climate change to cardiovascular health.

Extreme heat is increasingly recognized as a major risk factor for the global disease burden.⁶ Evidence from epidemiological studies suggests that exposure to extreme heat increases the risk of CVD mortality;^7,8,9 however, evidence regarding its impact on the incidence of CVD remains insufficient. Mechanistic studies suggest that extreme heat exposure could trigger systemic inflammation, oxidative stress, endothelial dysfunction, and cellular toxicity, potentially leading to CVD.^10,11 This suggests a potential association between extreme heat and CVD incidence. However, research outcomes on the relationship between extreme heat and CVD rates are mixed: some studies observe a correlation with increased CVD rates,^7,12 while others report minimal or no connection.^13,14,15 These inconsistencies underscore the complexity of determining the exact impact of extreme heat on CVD incidence, highlighting the need for further investigation, which may arise from diversity in exposure assessment and lag effects or co-exposure and interaction with other environmental stressors or heterogeneity in population vulnerability and adaptation capacity. Furthermore, current research predominantly focuses on short-term exposure to extreme heat, lacking longitudinal cohort studies based on prolonged exposure. Considering the significant implications of these studies for CVD prevention, intervention, and treatment, it is imperative to assess the effects of prolonged exposure to extreme heat on CVD incidence.

Exposure to fine particulate matter (PM_2.5) is associated with an increased risk of CVD incidence and mortality.^16,17 Both long-term and short-term exposure to PM2.5 have been confirmed to initiate a series of cardiovascular damage processes by inducing pulmonary and systemic inflammation and oxidative stress.¹⁸ Extreme heat and PM_2.5 share common underlying meteorological driving factors, often co-occur, and may synergistically amplify health impacts.¹⁹ More importantly, they may exert an impact on the cardiovascular system through similar pathophysiological pathways (such as systemic inflammation, oxidative stress, pro-thrombotic state, and endothelial dysfunction).²⁰ This overlap on the common biological pathway provides a reasonable theoretical assumption for the possibility of a synergistic effect between the two. To our knowledge, only two epidemiological studies have investigated the concurrent exposure to extreme heat and PM_2.5 on CVD mortality, revealing a synergistic effect that elevates the risk of CVD-related mortality.^9,21 Regarding the long-term combined effect of extreme heat and PM_2.5 on the incidence of CVDs and the specific mechanisms involved, there is still a lack of evidence from large-scale cohort studies. Particularly in recent decades, as climate change intensifies in low- to middle-income countries and air pollution concentrations continue to rise,²² significantly affecting the physical and mental health of individuals. Therefore, there is an urgent imperative to conduct a comprehensive investigation into the CVD risk arising from prolonged exposure to extreme heat and air pollution.

China, with its extensive territory, displays notable regional disparities in the sources, composition, concentrations, and spatiotemporal variations of climate and air pollutants, as well as population characteristics. In contrast to the southeastern coastal regions of China, the northwestern inland areas feature arid conditions, rendering them vulnerable to dust storms originating from the Taklamakan Desert in China and neighboring countries.^23,24 Moreover, the accelerated processes of industrialization and urbanization, coupled with the proliferation of arid climatic conditions, have resulted in a gradual escalation of air pollution levels and environmental temperatures in the northwestern region.^25,26 Given this context, there is an urgent imperative to thoroughly explore the combined impact of extreme heat and PM_2.5 on CVD incidence, which is essential for CVD prevention and alleviating its burden. This study, conducted from 2019 to 2021 in Xinjiang Uygur Autonomous Region, northwestern China, involved a cohort of 5.07 million adults. The aim was to assess the individual effects of extreme heat and PM_2.5 exposure on CVD incidence, identify susceptible populations, quantitatively analyze their interactions, and thereby determine the maximum risk threshold.

Results

Baseline characteristics

The basic characteristics of the study participants are presented in Table 1. Among the 5,070,689 participants without baseline CVD, 174,425 individuals developed CVD during the follow-up period. The mean age (standard deviation (SD)) and body mass index (BMI) (SD) of all participants were 42.79 (14.21) years and 24.97 (4.26) kg/m², with 55.88% being female. Compared to participants without CVD, those with CVD were older, mostly male, with higher BMI, a higher proportion of individuals with disabilities, and a greater number diagnosed with hypertension or diabetes. Median (interquartile range (IQR)) values for PM_2.5, annual ratio of extreme heat day (AREHD), and annual ratio of extreme heat hours (AREHHs) were 37.88 μg/m³ (27.86–45.23), 13.70 (6.85–16.44), and 4.25 (1.58–5.56), respectively (Table 2). The correlation between PM_2.5 and AREHD, as well as between PM_2.5 and AREHH, showed moderate correlations (rs of 0.530 and 0.551, respectively), while a high correlation was observed between AREHD and AREHH (rs of 0.971). It is worth noting that the baseline characteristics indicate that the proportion of “exercise frequency” among patients with CVD (8.99%) is higher than that among non-CVD populations (5.58%). This seemingly contradictory result may be related to the occupational sports activity bias. Studies show that higher occupational sports activities increase the risk of major adverse cardiovascular events (MACEs) and all-cause mortality.²⁷

Table 1.

Baseline characteristics of study participants

Variables	Total (n = 5,070,689)a	Non-CVD (n = 4,896,264)	CVD (n = 174,425)	p value
Age, years old	42.79 (14.21)	42.19 (13.94)	59.53 (11.41)	<0.001
Sex	–	–	–	<0.001
Male	2,237,089 (44.12)	2,159,759 (44.11)	77,330 (44.33)	–
Female	2,833,600 (55,88)	2,736,505 (55.89)	97,095 (55.67)	–
BMI, kg/m2	24.97 (4.26)	24.91 (4.23)	26.63 (4.61)	<0.001
Disability condition	–	–	–	<0.001
No disability	4,912,936 (96.89)	4,748,219 (96.98)	164,717 (94.43)	–
Physical disability	74,141 (1.46)	68,719 (1.40)	5,422 (3.11)	–
Other disabilities	83,612 (1.65)	79,326 (1.62)	4,286 (2.46)	–
Hypertension	–	–	–	<0.001
No	4,412,121 (87.01)	4,309,320 (88.01)	102,801 (58.94)	–
Yes	658,568 (12.99)	586,944 (11.99)	71,624 (41.06)	–
Diabetes	–	–	–	<0.001
No	4,804,728 (94.75)	4,660,242 (95.18)	144,486 (82.84)	–
Yes	265,961 (5.25)	236,022 (4.82)	29,939 (17.16)	–
Education level	–	–	–	–
Elementary school or below	1,893,395 (37.34)	1,790,569 (36.57)	102,826 (58.95)	<0.001
Middle school	2,320,137 (45.76)	2,272,661 (46,42)	47,476 (27.22)	–
High school	857,157 (16.90)	833,034 (17.01)	24,123 (13.83)	–
Marital status	–	–	–	<0.001
Single	453,410 (8.94)	451,620 (9.22)	1,790 (1.03)	–
Married	4,334,849 (85.49)	4,184,086 (85.45)	150,763 (86.43)	–
Divorced	103,264 (2.04)	100,780 (2.06)	2,484 (1.42)	–
Widowed	179,166 (3.53)	159,778 (3.26)	19,388 (11.12)	–
Family history of CVD	–	–	–	<0.001
No	4,462,394 (88.00)	4,309,381 (88.01)	153,013 (87,72)	–
Yes	608,295 (12.00)	586,883 (11.99)	21,412 (12.28)	–
Exercise frequency	–	–	–	–
Never	4,491,191 (90.54)	4,441,493 (90.71)	149,698 (85.82)	<0.001
Occasionally	143,904 (2.84)	137,812 (2.81)	6,092 (3.49)	–
More than once a week	46,495 (0.92)	43.542 (0.89)	2,953 (1.69)	–
Everyday	289,099 (5.70)	273,417 (5.58)	15,682 (8.99)	–
Eating habits	–	–	–	<0.001
Meat balanced	4,675,880 (92.21)	4,513,917 (92,19)	161,963 (92.86)	–
Meat based	200,567 (3.96)	194,563 (3.97)	6,004 (3.44)	–
Vegetarian based	194,242 (3.83)	187,784 (3.84)	6,458 (3.70)	–
Smoking status	–	–	–	<0.001
Never smoker	4,336,111 (85.51)	4,181,456 (85.40)	154,655 (88.67)	–
Current smoker	681,961 (13.45)	664,941 (13.58)	17,020 (9.76)	–
Former smoker	52,617 (1.04)	49,867 (1.02)	2,750 (1.58)	–
Alcohol drinking	–	–	–	<0.001
Never drinker	4,521,501 (89.17)	4,359,895 (89.05)	161,606 (92.65)	–
Occasional drinker	498,746 (9.84)	487,254 (9.95)	11,492 (6.59)	–
Regular drinker	42,663 (0.84)	41,720 (0.85)	943 (0.54)	–
Everyday drinkers	7,779 (0.15)	7,395 (0.15)	384 (0.22)	–

CVD, cardiovascular disease; BMI, body mass index.

Data are presented as mean (standard deviation) or n (%).

Table 2.

Descriptive statistics and correlation exploration of PM_2.5 and extreme heat

Descriptive statistics					Spearman correlation coefficients
Exposures	Mean±SD	Median (IQR)	Minimum	Maximum	PM2.5	AREHD	AREHH
PM2.5 (μg/m3)	37.47 ± 12.35	37.88 (27.86–45.23)	14.24	77.87	1.00	0.530 (<0.001)	0.551 (<0.001)
AREHD	12.11 ± 7.33	13.70 (6.85–16.44)	0	38.36	–	1.00	0.971 (<0.001)
AREHH	4.02 ± 3.16	4.25 (1.58–5.56)	0	20.19	–	–	1.00

IQR, interquartile range; PM_2.5, fine particulate matter; AREHD, annual ratio of extreme heat days; AREHH, annual ratio of extreme heat hours.

Effects of PM_2.5 and extreme heat on CVD

Long-term exposure to PM_2.5 and extreme heat significantly shortens the average time to onset of CVD (Tables 3 and S2–S4). In the fully adjusted models, for every 1 μg/m³ increase in annual PM_2.5 concentration, the average time to onset of CVD, coronary heart disease (CHD), and cerebrovascular disease decreases by 0.083% (−0.009–0.175), 0.378% (0.246–0.509), and 0.025% (−0.098–0.148), respectively. When AREHD increases by 1% (indicating 3.65 days per year exceeding 32°C), the average time to onset of CVD, CHD, and cerebrovascular disease shortens by 0.446% (0.304–0.587), 0.514% (0.312–0.715), and 0.470% (0.281–0.659), respectively. Similarly, for every 1% increase in AREHH (indicating 87.60 h per year exceeding 32°C), the average time to onset of CVD, CHD, and cerebrovascular disease decreases by 1.379% (1.043–1.714), 1.388% (0.918–1.855), and 1.499% (1.049–1.947), respectively.

Table 3.

Associations of PM_2.5 and extreme heat with CVD, CHD, and cerebrovascular onset time

Diseases (exposures)	Crude model		Fully adjusted model
	Time reduction ratio % (95% CI)	p value	Time reduction ratio % (95% CI)	p value
CVD
PM2.5	0.547 (0.456–0.638)	<0.001	0.083 (−0.009–0.175)	0.077
AREHD	0.672 (0.531–0.814)	<0.001	0.446 (0.304–0.587)	<0.001
AREHH	1.829 (1.492–2.165)	<0.001	1.379 (1.043–1.714)	<0.001
CHD
PM2.5	1.042 (0.911–1.172)	<0.001	0.378 (0.246–0.509)	<0.001
AREHD	0.790 (0.588–0.991)	<0.001	0.514 (0.312–0.715)	<0.001
AREHH	1.902 (1.432–2.369)	<0.001	1.388 (0.918–1.855)	<0.001
Cerebrovascular disease
PM2.5	0.426 (0.304–0.548)	<0.001	0.025 (−0.098–0.148)	0.688
AREHD	0.655 (0.465–0.844)	<0.001	0.470 (0.281–0.659)	<0.001
AREHH	1.878 (1.428–2.327)	<0.001	1.499 (1.049–1.947)	<0.001

CVD, cardiovascular disease; PM_2.5, fine particulate matter; AREHD, annual ratio of extreme heat days; AREHH, annual ratio of extreme heat hours; CHD, coronary heart disease.

Crude model: adjusted for age + sex.

Fully adjusted model: further adjusted for educational level, marital status, family history of CVD, exercise frequency, eating habits, smoking status, drinking frequency, body mass index, disability condition, hypertension, and diabetes mellitus.

In subgroup analyses, we observed the greatest effects of PM_2.5 exposure among the elderly (65 years old and above), men, individuals with a sedentary lifestyle, smokers, alcohol drinkers, persons without disability, and persons without hypertension or diabetes (Figure 1; Table S5). Additionally, we identified subgroups, including the elderly (≥65 years), males, and participants who never engage in physical activity, where the impact of extreme heat (AREHD, AREHH) is more pronounced. Furthermore, upon exposure to AREHD and AREHH, both smoking and alcohol drinking significantly shortened the time to the onset of CVD. Among alcohol drinkers, however, the effect of AREHH exposure was the strongest.

Figure 1.

Subgroup analysis of time reduction ratio for CVD associated with PM_2.5, AREHD, and AREHH

Abbreviations: CVD, cardiovascular disease; PM_2.5, fine particulate matter; AREHD, annual ratio of extreme heat days; AREHH, annual ratio of extreme heat hours. ∗, ∗∗, and ∗∗∗ indicated significance at the 0.05, 0.01, and 0.001 levels, respectively.

Tables 4 and S6 unveil the linear analysis of the interactive effects between PM_2.5 and extreme heat (AREHD, AREHH) on CVD. We observed interactions between AREHD and AREHH with PM_2.5, with time reduction ratios of −0.03 (95% CI: −0.06–−0.00) and −0.03 (95% CI: −0.10–−0.04), respectively. The analysis of the impact of PM_2.5 and extreme heat on the average onset time of CVDs in different exposure combinations revealed that PM_2.5 exposure was the primary factor influencing the onset time (Figure 2). No significant changes in the onset time of CVDs were observed under low PM_2.5 exposure levels, regardless of high or low AREHD or AREHH. Compared with the reference group exposed to both low PM_2.5 and low heat, simultaneous exposure to moderate-to-high PM_2.5 and extreme heat, particularly high AREHH, was associated with a significant reduction in the onset time of CVDs (PM_2.5 and AREHD: time reduction ratio, 11.97; 95% CI, 8.25–15.55; PM_2.5 and AREHH: time reduction ratio, 11.30; 95% CI, 7.74–14.72).

Table 4.

Interactive effects of PM_2.5 and extreme heat on CVD

Modelsa	PM2.5	Extreme heat	Multiplicative interaction between PM2.5 and extreme heat
PM2.5 + AREHD + PM2.5 × AREHD	1.07 (0.75–1.39)∗∗	0.86 (0.04–1.67)∗	−0.03 (−0.06–−0.00)∗
PM2.5 + AREHH + PM2.5 × AREHH	0.81 (0.52–1.09)∗∗	1.19 (−1.23–3.56)	−0.03 (−0.10–−0.04)

CVD, cardiovascular disease; PM_2.5, fine particulate matter; AREHD, annual ratio of extreme heat days; AREHH, annual ratio of extreme heat hours.

∗, ∗∗, and ∗∗∗ indicated significance at the 0.05, 0.01, and 0.001 levels, respectively.

^aModels were adjusted for age, sex, educational level, marital status, family history of CVD, exercise frequency, eating habits, smoking status, drinking frequency, body mass index, disability condition, hypertension, and diabetes mellitus.

Figure 2.

The influence of different subgroups, based on PM_2.5 and extreme heat (AREHD, AREHH), on CVD onset time

The time reduction ratio for CVD, stratified by (A) AREHD, (B) AREHH, and PM_2.5, was estimated using AFT models with adjustments for age, sex, educational level, marital status, family history of CVD, exercise frequency, eating habits, smoking status, drinking frequency, body mass index, disability condition, hypertension, and diabetes mellitus.

Abbreviations: CVD, cardiovascular disease; PM_2.5, fine particulate matter; AREHD, annual ratio of extreme heat days; AREHH, annual ratio of extreme heat hours; AFT, accelerated failure time.

Dose-response of PM_2.5 and extreme heat on CVD

To assess dose-response relationships and potential non-linearities, we generated bivariate dose-response surfaces for the combined effects of PM_2.5 concentration and extreme heat (AREHD, AREHH) on CVD onset time (Figure 3: A & D). Based on a large community cohort in Xinjiang - Tianshan Community Cohort Establishment and Follow-up, this study employed an accelerated failure time (AFT) model with restricted cubic splines to unveil a complex nonlinear dose-response relationship between exposures and CVD risk.

Figure 3.

Dose-response relationship of PM_2.5 exposure and extreme heat (AREHD, AREHH) on the CVD onset time

(A and D): Dose-response surface of annual PM_2.5 concentration, AREHD, AREHH, and CVD onset time.

(B and E): The effect of annual PM_2.5 concentration on CVD onset time by different AREHD or AREHH levels.

(C and F): The effect of AREHD or AREHH on CVD onset time by different PM_2.5 levels.

Abbreviations: CVD, cardiovascular disease; PM_2.5, fine particulate matter; AREHD, annual ratio of extreme heat days; AREHH, annual ratio of extreme heat hours.

Figure 3A, a three-dimensional surface plot, demonstrates that under the joint exposure of AREHD and PM_2.5, the predicted time to CVD onset exhibits a distinct “valley-shaped” distribution. The shortest time to onset—indicating the highest risk—occurs in the medium exposure range, approximately at a PM_2.5 concentration of 32.7 μg/m³ and an AREHD of about 10%. However, in higher exposure regions (e.g., PM_2.5 > 45 μg/m³ and AREHD >15%), the time to onset unexpectedly lengthens, suggesting potential model extrapolation uncertainty or a “harvesting effect” among high-risk populations. Figure 3B further confirms the nonlinearity of the PM_2.5 effect, the shape of which is modified by AREHD levels. At low AREHD (1.37%), PM_2.5 shows a clear “U-shaped” relationship with the time to onset, with the peak risk (shortest time to onset of 149.1 years) occurring at a PM_2.5 concentration of 32.71 μg/m³. As AREHD levels increase (e.g., to 9.59% and 15.62%), the U-shaped curve flattens and the minimum point shifts rightward, indicating that high-temperature exposure attenuates the independent effect of PM_2.5. Figure 3C reveals that the direction of the AREHD effect depends on the background PM_2.5 level. Under low PM_2.5 conditions (22.92 μg/m³), an increase in AREHD is associated with a slight prolongation of the time to onset. In contrast, under medium-high (42.43 μg/m³) to high (48.36 μg/m³) PM_2.5 conditions, an increase in AREHD leads to a significant shortening of the time to onset.

Comparing the two extreme heat indicators, Figures 3D–3F demonstrate that AREHH is a stronger risk factor and effect modifier. Its three-dimensional surface (Figure 3D) is steeper, and Figure 3F shows that against a high PM_2.5 background (48.36 μg/m³), an increase in AREHH from 0% to 6.1% results in a drastic shortening of the predicted time to onset, with a hazard magnitude far exceeding that of AREHD. Furthermore, in Figure 3E, the U-shaped effect of PM_2.5 under low AREHH (0.16%)—with a minimum time to onset of 148.5 years at PM_2.5=34.13 μg/m³—is also more pronounced than in the AREHD model. In summary, PM_2.5 and extreme heat exhibit a complex nonlinear joint effect on CVD risk, characterized by a “peak risk window” at medium exposure levels.

Sensitivity analysis

Sensitivity analyses demonstrate the robustness of our model results. Whether analyzing the data without imputation (Table S7), examining two- and three-year average exposures of PM_2.5 and extreme heat (AREHD, AREHH) (Table S8), or employing Cox proportional hazards models (Tables S9 and S10), or adjusting for several covariates in the model (Table S11), or include the 85th and 90th percentile-based temperature events (Table S12), the findings remain mainly consistent with our primary analysis.

Discussion

In this cohort study conducted in 2019, we utilized the AFT model to explore the association between PM_2.5 and extreme heat with the CVD incidence. We discovered that long-term exposure to PM_2.5 and extreme heat significantly increases the risk of developing CVD. Subgroup analysis revealed a notable elevation in PM_2.5-induced CVD risk among the elderly (>=65 years), male, physically inactive, smoking, alcohol-consuming, non-disabled individuals, and those with diabetes or hypertension. Moreover, in the elderly (≥65 years), males, and non-exercisers, the association between extreme heat (AREHD, AREHH) and CVD was more pronounced. The bivariate dose-response surface and stratified analyses demonstrate that PM_2.5 and extreme heat mutually amplify their detrimental effects, leading to the greatest reduction in CVD survival time under concurrent medium exposure to both factors.

To our knowledge, multiple epidemiological reports have indicated that exposure to PM_2.5 increases the risk of CVD incidence.^28,29 However, these pieces of evidence largely originate from countries or regions with relatively lower levels of air pollution.^30,31 Household activities, agriculture, power generation, and land transportation constitute the primary sources of PM_2.5 pollution in these areas. Hence, this may limit the applicability of the study findings to regions characterized by severe pollution dominated by natural dust. Our research findings regarding the impact of PM_2.5 on the incidence of CVD fill this gap. The observed association between PM_2.5 and CVD in our study is consistent with other research, although the degree of association varies.^31,32 For instance, concerning CVD incidence, our estimation indicates a hazard ratio (HR) of 1.00117 (95% CI: 1.00023–1.00212) for each 1μg/m³ increase in PM_2.5, while Xia et al. reported an HR of 1.12 (95% CI: 1.11–1.14).³¹ Although the proportional hazards assumption was not fully satisfied in the primary model, the results from the Cox model still offer a meaningful reference value. For the incidence of CHD and cerebrovascular disease, our estimates for each 1 μg/m³ increase in PM_2.5 yielded Time Reduction Ratios of 0.378 (95% CI: 0.246–0.509) and 0.025 (95% CI: −0.098–0.148), respectively, whereas Wei et al. reported that the time ratios (TRs) of stroke associated with PM₂.₅ was 0.958 (95% CI: 0.953–0.964), which also indicates a risk factor.³³ The differences in these study outcomes may be attributed to various factors such as diverse sources of PM_2.5 in different regions, variations in inclusion/exclusion criteria, disease definitions, and follow-up durations among the studies. The subgroup analysis of PM_2.5 exposure across different populations aligns generally with previous reports. Specifically, we observed a stronger correlation in males and those with unhealthy lifestyles (no exercise, smoking, and alcohol consumption), consistent with earlier studies.^31,34,35 Additionally, we found that smoking and alcohol consumption significantly amplified the harmful effects of both extreme heat indicators, showing higher Time Reduction Ratios and greater risk magnitudes compared to other subgroups, which could be considered among the most critical risk factors. Potential explanations include that alcohol can lead to dehydration, and extreme heat exacerbates this condition, jointly increasing blood viscosity and thereby elevating cardiac load, which synergistically accelerates cardiovascular damage. Meanwhile, nicotine in tobacco may harm atrial myocardium through shared mechanisms such as oxidative stress, inflammatory responses, atrial fibrosis, and autonomic imbalance, both of which substantially impact the cardiovascular system.³⁶

Our study results on the association between exposure to extreme heat and the risk of CVD incidence support and complement previous research. Specifically, Liu et al. conducted a meta-analysis focusing on heat exposure and CVD, finding a significant increase in the incidence of arrhythmias, cardiac arrest, and CHD associated with a 1°C rise in temperature.⁷ Tao et al. further explored the impact of daytime and nighttime high temperatures on the risk of CVD incidence, demonstrating that both daytime and nighttime high temperatures elevate the risk of CVD, with nighttime high temperatures associated with a higher risk than daytime.⁸ Our study, taking a long-term exposure perspective, addresses gaps in previous research, contributing to the understanding of this field. Regarding the impact of extreme heat on different subgroups, we observed a higher level of vulnerability in the elderly population (≥65 years), aligning with previous studies that partially explored CVD mortality.^21,37 This may be attributed to differential physiological and pathological responses to heat stress in the elderly,³⁸ along with variations in social and living conditions.³⁹ Notably, we also observed a closer association between extreme heat and the risk of CVD incidence in males and individuals with a sedentary lifestyle, providing additional insights into potential susceptibility factors.

The findings of this study regarding the synergistic effect of extreme heat and PM_2.5 on the incidence of CVDs are consistent with the recent research conclusions based on mortality data, further validating the robustness and reliability of the study results, and supporting that both extreme heat and PM_2.5 are independent and interacting risk factors for CVDs. For instance, the population study results of Xu et al. (2023) suggest that dual exposure poses a greater threat to cardiovascular health than single exposure. Based on the time-stratified case-crossover design, it was discovered that heat waves and PM_2.5 exposure have a significant synergistic interaction on the risk of myocardial infarction mortality.²¹ Our study not only confirmed the existence of this synergistic effect in a large prospective cohort but also further revealed its complex nonlinear nature. This difference may stem from differences in study outcomes, study designs, and exposure assessment methods. This finding suggests that when assessing the health risks of environmental combined exposures, special attention should be paid to the nonlinear dose-response relationship. The co-occurrence of extreme heat and poor air pollution has become more frequent in recent years.^9,20 While evidence on the interactive impact of extreme heat and PM_2.5 on CVD mortality is gradually strengthening,^9,21 the potential synergistic effects of these exposures on CVD survival have not been fully elucidated. To our knowledge, this is the first systematic investigation into the long-term interactive effects of extreme heat and PM_2.5 on survival time among patients with CVD. Our groundbreaking study indicates a complex synergistic effect between PM_2.5 and extreme heat on CVD survival time.

The stratified analysis of the relationship between PM_2.5 exposure and CVD onset time under different exposure combinations with extreme heat further revealed the primary-secondary relationship in risk transmission: PM_2.5 exposure is the dominant factor influencing onset time. At low PM_2.5 exposure levels, no significant changes in CVD onset time were observed regardless of AREHD or AREHH levels, suggesting that under a low pollution background, the independent pathogenic effects of moderate-intensity heat may be buffered by population baseline adaptive capacity. However, when medium-to-high PM_2.5 concentrations co-exposed with extreme heat (particularly high AREHH), the CVD onset time was significantly shortened, with time reduction ratios reaching 11.97% and 11.30%, respectively. This indicates a clear threshold-dependent synergistic amplification effect between PM_2.5 and extreme heat—where the health hazards of heat require a certain pollution load to be fully activated, while the cardiovascular toxicity of PM_2.5 is further intensified under high-temperature conditions. The findings indicate that their interaction exhibits significant nonlinear characteristics, with the peak risk occurring not at the highest exposure levels but within a moderate exposure range. This is different from most studies that report on the linear synergistic effect (such as those by XU,²¹ Rahman, etc.⁴⁰) Cardiovascular survival time is most markedly shortened (peak risk) under concurrent exposure to PM_2.5 concentrations around 32.7 μg/m³and extreme heat day proportion (AREHD) around 10%. Stratified analyses further confirm this dynamic interaction: under low-level heat exposure, PM_2.5 shows a “U-shaped” dose-response relationship with survival time; as heat levels increase, this curve flattens and the risk peak shifts rightward. Simultaneously, the direction of the hazard effect from extreme heat depends on the background PM_2.5 level—under lower pollution conditions, the independent effect of heat is limited, but against a medium-to-high PM_2.5 background (e.g., >40 μg/m³), heat exposure significantly exacerbates cardiovascular risk. Particularly noteworthy is that the extreme heat indicator defined by hourly proportion (AREHH) demonstrates a stronger capacity to modify risk. Against a high PM_2.5 background, even minor increases in AREHH can lead to a sharp decline in cardiovascular survival time, suggesting that sustained heat stress poses an independent hazard that cannot be ignored. Against the backdrop of intensifying global heatwaves and regional improvements in air pollution levels, these findings outline a new public health risk landscape: the greatest threat to cardiovascular health may not stem from high exposure to a single environmental factor, but rather resides within a specific “synergistic window” of moderate pollution and moderate heat. This insight holds significant implications for developing precise early-warning standards for combined exposures and implementing targeted public health interventions, the interplay of these factors becomes increasingly critical, especially for China with its climate diversity and aging population.⁴¹ The overlap of extreme heat and air pollution is a pressing challenge. Our research, by addressing this intersection, offers meaningful insights into public health planning and preventive measures, underscoring the importance and timeliness of our findings in the face of evolving global and national environmental health challenges.

Although the underlying biological mechanisms of the deep-seated relationship between PM_2.5 and CVD have not been fully elucidated, various potential mechanisms have been proposed. The synergistic damage to the cardiovascular system caused by extreme heat and PM_2.5 is based on the fact that both share multiple core pathological and physiological pathways.²⁰ Firstly, oxidative stress and systemic inflammation are the key common initiating mechanisms. After inhalation of PM_2.5, it can induce the production of reactive oxygen species in the lungs, triggering local and systemic inflammatory responses¹⁸; while extreme exposure to extreme heat itself can directly trigger a systemic inflammatory cascade reaction.⁴² When both exist together, the levels of inflammation and oxidative stress may be superimposed or synergistically amplified. Secondly, both can cause endothelial dysfunction. PM_2.5 directly damages the endothelium by reducing the bioavailability of nitric oxide,¹⁸ and the inflammation and hemodynamic changes caused by extreme heat may also damage the vasodilatory function. Thirdly, the pro-thrombotic state is another important pathway. PM_2.5 exposure can activate platelets, increasing blood coagulability¹⁸; extreme heat also leads to dehydration and blood concentration, increasing the risk of thrombosis.⁴² Moreover, autonomic nerve function imbalance and biological rhythm disorders are also considered as potential pathways for the joint influence of both.²⁰ When an individual is simultaneously exposed to extreme heat and PM_2.5, these overlapping pathways may be simultaneously activated and interact with each other, thereby generating a pathogenic effect beyond the effects of a single exposure, which provides a mechanism explanation for the observed synergistic health risks.

It is worth noting that this study found that the peak of combined risk occurred at a moderate exposure level, rather than the highest exposure level, this reveals a non-linear interaction pattern. We hypothesize that the following mechanisms might explain this: 1. Biological pathway saturation effect. Both PM_2.5 and extreme heat can activate common pathological physiological pathways, including oxidative stress, systemic inflammatory response, and autonomic nervous system dysfunction. When both factors are at a moderate level, these pathways are simultaneously activated, resulting in a synergistic amplification effect. However, when the exposure level exceeds a certain threshold, these pathways may become saturated, and at this point, the additional stimulus from extreme heat cannot further amplify the already maximized pathological response. 2. Adaptive compensatory response. Long-term high-level exposure to PM_2.5 may induce the body’s adaptive mechanisms, such as the upregulation of heat shock proteins and the activation of the antioxidant defense system, which to some extent buffer the acute damage effects of the combination of high temperature and high pollution. In a study published by Eric S. Coker et al. in the journal ⟪npj Clean Air⟫, it was also observed that the mortality risk was highest when the moderate exposure level of PM_2.5 (12–14 μg/m³) was combined with high temperature (≥26°C). And in their paper, they gave the following explanations: First, the occurrence of biological saturation phenomenon; second, behavioral change - during high pollution, people tend to stay indoors and close the doors and windows, resulting in a reduction in actual exposure, which underestimates the high exposure effect; third, data sparsity - the number of days when extreme high temperature (≥26°C) and extreme PM_2.5 (≥16 μg/m³) occur simultaneously is only 1.2%, and the statistical power is insufficient.⁴³ This non-linear synergistic pattern emphasizes that in the context of climate change and air pollution control, public health concerns should not be limited to extreme exposure events, but the risk warning of moderate-level combined exposure is also crucial. These mechanisms jointly explain the synergistic damage effect of high temperature and PM_2.5 on the cardiovascular system, providing a solid biological basis for the observed epidemiological association in this study.

Our study has several strengths. First, by including Xinjiang, a region with unique geographical conditions and differences in population lifestyle, the external validity and regional specificity of the conclusion have been significantly enhanced. Second, breaking through the limitations of the traditional single-exposure model, a composite exposure index system has been adopted to more accurately reflect the real exposure scenarios where multiple environmental factors interact synergistically, aligning with the methodological frontier in public health that moves from single-factor analysis to multi-factor integrated analysis.

In conclusion, this study provides additional evidence indicating a significant association between long-term exposure to extreme heat and PM_2.5 and an increased risk of developing CVD. Our analysis identifies that the highest risk occurs in the medium exposure range. This study provides valuable insights for preventing, intervening, and treating CVD by considering the independent and interactive effects of extreme heat and PM_2.5, particularly for high-risk populations and clinicians.

Moreover, these findings offer significant implications for public health policy. The study identifies key susceptible groups—such as sedentary individuals, smokers, drinkers, and the elderly—providing a basis for targeted health protection. We recommend that Xinjiang establish a pilot hierarchical early warning mechanism for composite exposure to high temperature PM_2.5 based on existing single pollutant alerts, targeting the risk window of PM_2.5 concentrations around 32–35 μg/m³ combined with approximately 10% of high temperature days, during which tailored cardiovascular health protection guidelines should be disseminated to the public. However, owing to the unique climatic conditions of Xinjiang (e.g., high altitude, low temperatures, aridity), the study findings are derived from a region-specific context and may not be broadly generalizable worldwide. Nonetheless, as a core region of the belt and road initiative (BRI), Xinjiang’s environmental and population characteristics are comparable to those of many BRI countries, making our results a valuable reference for cardiovascular risk management in those nations.

Limitations of the study

This study has several limitations. First, the environmental exposure levels for all participants were based on the baseline year of the cohort (2019), which may introduce potential misclassification of exposure. However, we observed that air pollutant levels from 2017 to 2019 were stable, thus the potential impact of misclassification would be minimal. Second, similar to many previous cohort studies, we relied on environmental exposures based on residential addresses, without considering issues related to time-location activities. Thirdly, despite efforts to control for confounding factors, residual confounding cannot be completely ruled out. For instance, participants from more affluent households may be more likely to reside in areas with lower environmental exposure levels, potentially reducing the risk of CVD incidence. Finally, the lack of data on household air pollution may introduce residual confounding, as household air pollution sources are often correlated with outdoor pollution levels and socio-economic conditions.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Kuiying Gu (kygu@suda.edu.cn).

Materials availability

The authors confirm that the materials are available from the lead contact upon reasonable request.

Data and code availability

• •Data: Data will be made available on request.
• •Code: This paper does not report the original code.
• •Additional information: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by the Suzhou Applied Basic Research (Medical and Health) Science and Technology Innovation Project (no. SYW2025193).

Author contributions

Conceptualization, K.G., L.K., and Y.Y.; methodology, W.Z., Y.W., Q.X., J.O., Q.D., and Y.Y.; software, W.Z., Y.W., Q.X., J.O., and Q.D.; validation, J.O., Q.D., Y.Y., Y.Y., K.G., and L.K.; formal analysis, W.Z., Y.W., Q.X., J.O., and Q.D.; investigation, W.Z., Y.W., and Q.X.; resources, K.G. and L.K.; data curation, W.Z., Y.W., and Q.X.; visualization, W.Z., Y.W., and Q.X.; supervision, K.G. and L.K.; project administration, K.G. and L.K.; funding acquisition, K.G.; writing – original draft, W.Z., Y.W., and Q.X.; writing – review and editing, W.Z., Y.W., Q.X., J.O., Q.D., Y.Y., Y.Y., K.G., and L.K.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Tianshan Community Cohort	Tianshan Community Cohort Establishment and Follow-up	https://clinicaltrials.gov/
PM2.5	The Tracking Air Pollution (TAP) dataset	http://tapdata.org.cn
Temperature	The European Center for Medium-Range Weather Forecasts atmospheric reanalysis v5 land (ERA5-land) dataset	https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land?tab=overview
Software and algorithms
R version 4.5.1	R Foundation	R: The R Project for Statistical Computing

Experimental model and study participant details

A longitudinal cohort study was conducted by establishing a cohort using community health examination data from residents of 14 prefectures and cities in Xinjiang Uygur Autonomous Region, Northwestern China. Approximately 10 million individuals participate in annual community health examinations, which include questionnaire surveys, routine physical examinations, and laboratory tests. Routine examinations include measuring standing height, weight, waist circumference, blood pressure, electrocardiogram, ultrasound, and X-ray examinations, among others. Laboratory tests include analyses of blood glucose and biochemical markers. Examinations are performed by qualified and professional medical personnel. Participants must carry their unique identification card, which serves as the only form of identification.

This prospective cohort study analyzed individual-level community health examination data from residents in 2019, 2020, and 2021. The cohort, initially consisting of 6,179,679 participants identified by their ID cards, excluded individuals meeting the following criteria: (1) diagnosed with CVD at baseline (n= 953,329), (2) missing important information such as height, weight and sex. (n= 33,740), (3) with any occurrences of CVD prior to baseline (n=34,681), and (4) inability to match PM_2.5 or environmental concerns data with the place of residence and year of follow-up (n=87,060). Finally, 5,070,689 participants were included in the analysis (Figure S1).

All participants underwent concurrent follow-up, with dynamic surveillance conducted based on hospitalization records and subsequent routine health examinations. The study endpoint was defined as the first occurrence of CVD, CVD-related death or the follow-up cutoff date of December 31, 2021, whichever occurred first. The median follow-up time of this cohort was 2.57 years, with a mean follow-up time of 2.54 (± 0.30) years.

Our cohort was established based on the Tianshan Community Cohort (Tianshan Community Cohort Establishment and Follow-up). This study protocol has been successfully registered on the international clinical trial registry platform (ClinicalTrials.gov) with registration number NCT07207863.The study received approval from the Medical Ethics Committee of Tsinghua University and required written informed consent from all participants undergoing health examinations. This study used de-identified data from community health examinations and thus informed consent of individuals was not required.

Method details

Outcome assessment

The “timing of first CVD onset” is defined as the date of first confirmed CVD diagnosis recorded in the electronic medical record data (2019–2021). And physician coding of cardiovascular events is based on the International Classification of Diseases, Tenth Revision (ICD-10). Initially, CVD history was assessed through five questions posed by medical staff to participants. Subsequently, physicians further defined cardiovascular events around the ICD-10 based on the examination results: rheumatic heart disease as I01 and I05-I09, hypertensive heart disease as I11, ischemic heart disease as I20-I25, pulmonary embolism and other acute pulmonary heart diseases as I26-I28, other forms of heart disease as I34-I36, I42-I43, and I50-I51, and cerebrovascular diseases as I60-I67 and I69.

Exposure assessment

The annual ambient concentrations of PM_2.5 (spatial resolution of 10 km × 10 km) were derived from the Tracking Air Pollution (TAP) dataset.⁴⁴ Temperature data for this study were sourced from the European Centre for Medium-Range Weather Forecasts atmospheric reanalysis v5 land (ERA5-Land).⁴⁵ We calculated the annual average PM_2.5 exposure and hourly temperature data for each participant based on their residence's latitude and longitude from 2017 to 2019 (Figure S2B). Prior research has confirmed that temperatures exceeding 32°C have detrimental effects on human health.^46,47,48 In this study, we defined two measures for exposure to extreme heat: the annual ratio of extreme heat days (AREHD) and the annual ratio of extreme heat hours (AREHH) (Figures S2C and S2D). AREHD represents the percentage of days in a year with temperatures above 32°C (Days/365×100%), while AREHH represents the percentage of hours within a day with temperatures exceeding 32°C (Hours/24/365×100%).

Covariates assessment

We extracted covariate information from the community health examination data, including age (years), sex (male, female), weight, height, disability condition (no disability, physical disability, other disabilities), hypertension (no, yes), diabetes (no, yes), educational level (elementary school or below, middle school, high school), marital status (single, married, divorced, widowed, not specified), family history of CVD (no, yes), exercise frequency (never, occasionally, more than once a week, everyday), smoking status (never smoker, former or current smoker), alcohol drinking (never drinker, former or current drinker), and eating habits (meat balanced, meat based, vegetarian based). Body mass index (BMI) was calculated by dividing weight (in kilograms) by height (in meters) squared. The status of missing values for each covariate is detailed in Table S1. To mitigate cross-regional influences, we applied the random forest method separately in each region for imputing missing values.⁴⁹

Quantification and statistical analysis

We first assessed the baseline differences in characteristics between participants with and without CVD. Due to the non-satisfaction of the proportional hazards assumption for the covariates under analysis, we utilized an accelerated failure time model (AFT) instead of the Cox proportional hazards model to estimate the association between PM_2.5, extreme heat, and the health outcomes of interest. We established two models: the crude model, adjusting for age and sex only, and the fully adjusted model, accounting for all covariates. All AFT models were stratified by district/county of residence to account for unobserved heterogeneity at the regional level. In each model, we reported the mean time reduction ratio of health outcomes when PM_2.5 increased by 1 μg/m³ or when AREHD (or AREHH) increased by 1%, along with the corresponding 95% confidence intervals (CIs) (the formula for calculating the time reduction ratio is (1-exp(estimated coefficients))×100%). To further identify sensitive subgroups regarding PM_2.5 and extreme heat, we conducted subgroup analyses based on age (<65 years, ≥65 years), sex (male vs. female), exercise frequency (no exercise, exercise), smoking status (never smoker, former or current smoker), alcohol drinking (never drinker, former or current drinker), disability condition (no, yes), hypertension (no, yes), and diabetes (no, yes). We examined potential effect modifications using 2-sample Z-tests, comparing the stratification-specific point estimates (β = ln odds ratio) and their standard errors (SEs).⁵⁰

To assess the combined impact of PM_2.5 and extreme heat on CVD, we introduced interaction terms for PM_2.5 with AREHD or AREHH in the primary analysis. We explored the effects of different PM_2.5 levels and AREHD (or AREHH) on CVD. Participants were grouped based on quantile values: low, intermediate, and high PM_2.5 exposure (at 33.3% and 66.6% quantiles, i.e., 29.67μg/m³ and 36.79 μg/m³), respectively. Additionally, using the median value of AREHD (or AREHH) at 11.51 (or 3.38), participants were categorized into low or high AREHD exposure (or low or high AREHH exposure levels).

To explore potential non-linearity, we introduced a restricted cubic spline with four nodes for PM_2.5 and extreme heat (AREHD, AREHH) in the model with interaction terms. To mitigate the influence of extreme values, we limited the study subjects to those exposed to PM_2.5 concentrations and AREHD (or AREHH) unit values within the 7.5%-92.5% range. We visualized the three-dimensional relationship among PM_2.5 exposure, extreme heat (AREHD, AREHH) exposure, and the risk of CVD incidence. Additionally, we separately assessed the dose-response relationship between PM_2.5 and CVD risk at different exposure levels of extreme heat and the impact of extreme heat on the risk of CVD incidence at varying levels of PM_2.5.

To ensure robustness, we conducted sensitivity analyses. First, we analyze the impact of PM_2.5 exposure and extreme heat on CVD using a dataset with over 4.9 million samples. This dataset was not subjected to random forest imputation, and samples with missing covariate values were excluded. Then, to further strengthen the robustness of our findings, we excluded participants with physical disabilities in the following analysis, given their limited outdoor activity and correspondingly lower outdoor exposure. Specifically, these analyses include (1) testing for potential exposure misclassification by utilizing two- and three-year average exposures of PM_2.5 and extreme heat (AREHD, AREHH), (2) reanalyzing the data using Cox proportional hazards models to evaluate the robustness of the study results, (3) reanalyzing the data by adjusting for several covariates in the model, (4) including 85th and 90th percentile-based temperature events by using two-year exposure windows.

All analyses were conducted using the “survival” package in R version 4.5.1 statistical software. A significance level of two-tailed P < 0.05 was used to determine statistical significance.

Additional resources

This cohort study protocol has been successfully registered on the international clinical trial registry platform (ClinicalTrials.gov) with registration number NCT07207863.