Long-term exposure to ambient air pollution and incident cardiovascular disease in adults with cardiovascular–kidney–metabolic stages 0–3: a nationwide prospective cohort study in China Authors: Wei Li, Fangfang Lou, Xiaoyi Liu, Wenfeng Ye, Yunshi Lai, Zijie Liang, Huishi Li, Xuefang Huang, Peiyi Ye, Yaozhong Kong, Chao Xie Published: 2025 ABSTRACT AbstractCardiovascular–Kidney–Metabolic (CKM) stages 0–3 represent a key window for cardiovascular disease (CVD) prevention, yet the effects of long-term ambient air pollution in this population remain unclear. In this longitudinal analysis of a nationwide prospective cohort, we analyzed 6,321 adults without baseline CVD from the China Health and Retirement Longitudinal Study (CHARLS). Long-term exposure to PM2.5, NO2, and warm-season O3 was estimated using the China High Air Pollutants satellite-based dataset, which provides pollutant surfaces at 1-km resolution (10-km for NO₂). Incident CVD through 2018 was physician-diagnosed. Cox models were used to evaluate baseline and time-varying associations, and non-linear and joint effects were examined using spline and multipollutant approaches. During a median follow-up of 6.9 years, 896 participants developed CVD. Each 10 µg/m³ increase in baseline PM2.5 or NO2 was associated with a 20% higher CVD risk (PM2.5: HR 1.20, 95% CI 1.13–1.28; NO2: HR 1.20, 1.07–1.35). Time-varying models confirmed these findings and additionally identified a significant association with O3 (HR = 1.20, 95% CI: 1.11–1.30). Restricted cubic spline analyses indicated non-linear exposure–response relationships, with thresholds of 39.64 µg/m3 for PM2.5 and 39.86 µg/m3 for NO2. Multipollutant models showed a positive mixture effect dominated by PM2.5, and stronger associations were observed in participants from eastern China (P interaction < 0.01). Long-term exposure to PM2.5, NO2, and O3 increases CVD risk in adults with CKM stages 0–3, with PM2.5 as the leading driver. These findings highlight the need to incorporate air pollution into CKM-based cardiovascular prevention strategies.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-32546-1.Keywords: Air pollution, Cardiovascular diseases, Cardiovascular–kidney–metabolic syndrome, PM2.5Subject terms: Cardiology, Diseases, Environmental sciences, Risk factors Introduction The Cardiovascular-Kidney-Metabolic (CKM) syndrome, introduced by the American Heart Association (AHA) in 2023, reframes cardiovascular disease (CVD), chronic kidney disease (CKD), type 2 diabetes, and obesity into a single progressive disorder. The early CKM stages (0–3) represent a critical window for prevention before overt CVD develops1. This framework is particularly relevant in China, which bears a massive and poorly controlled burden of CKM components, including the world’s largest populations with hypertension and diabetes2–5. Current AHA guidelines for CKM management emphasize lifestyle modifications, pharmacotherapies, and social determinants of health (SDOH)1,6. While these include factors like healthcare access and income, a critical SDOH—ambient air pollution—is notably absent7. This omission is significant, as pollutants such as fine particulate matter (PM2.5), nitrogen dioxide (NO2), and ozone (O3) promote the same biological pathways—systemic inflammation, oxidative stress, and endothelial dysfunction—that drive CKM progression8–10. The oversight is particularly hazardous in China, where the 2024 national average PM2.5 concentration (29.3 µg/m³) remains nearly six times the World Health Organization (WHO) guideline, subjecting millions to chronic high-level exposures11. Although previous studies have linked air pollution to individual CKM components12–16, no research has quantified whether chronic exposure accelerates the progression to incident CVD specifically among the millions of individuals in early CKM stages (0–3). We therefore aimed to determine if long-term exposure to PM2.5, NO2, and O3 increases CVD risk in this vulnerable population, providing evidence to support the integration of environmental and clinical prevention strategies. Method Study population This study utilized data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative prospective longitudinal survey targeting Chinese residents aged 45 years and older. The baseline survey was conducted between 2011 and 2012, encompassing approximately 17,705 individuals from 10,257 households across 150 counties/districts and 450 villages/urban communities in 28 provinces. Follow-up waves were administered biennially in 2013 (Wave 2), 2015 (Wave 3), 2018 (Wave 4), and 2020 (Wave 5). Detailed information regarding the study population has been previously published. The CHARLS protocol was approved by the Institutional Review Board of Peking University (IRB00001052-11015), and all participants provided written informed consent prior to participation17. For the present analysis, we included individuals who participated in the 2011 baseline and had at least one follow-up before 2018 (Wave 4). The flow diagram (Fig. 1) presents the inclusion and exclusion criteria for this study. We excluded individuals who had missing data required for CKM staging (n = 7,732), who reported baseline CVD (heart disease or stroke) (n = 2,956), or who were younger than 45 years at baseline (n = 214) or did not participate in any follow-up waves (n = 482). Flowchart of study population selection. Participants were drawn from the 2011 baseline wave of the CHARLS. Individuals were excluded if they had baseline CVD, were under 45 years of age, had missing data for CKM staging, or were lost to follow-up. Data assessment Assessment of CKM syndrome stages 0–3 CKM syndrome stages were defined in accordance with the AHA’s 2023 Presidential Advisory Statement1. At baseline (Wave 1), participants were categorized as follows: Stage 0: Individuals without overweight/obesity, metabolic risk factors, CKD, or subclinical/clinical CVD. Stage 0: Individuals without overweight/obesity, metabolic risk factors, CKD, or subclinical/clinical CVD. Stage 1: Presence of excess adiposity (overweight/obesity or abdominal obesity) and/or dysfunctional adipose tissue, without CKD or other metabolic risk factors. Stage 1: Presence of excess adiposity (overweight/obesity or abdominal obesity) and/or dysfunctional adipose tissue, without CKD or other metabolic risk factors. Stage 2: Presence of metabolic risk factors (e.g., hypertension, type 2 diabetes), moderate-to-high-risk CKD, or both. Stage 2: Presence of metabolic risk factors (e.g., hypertension, type 2 diabetes), moderate-to-high-risk CKD, or both. Stage 3: Subclinical CVD. In the absence of direct biomarkers, this stage was operationally defined by CKM risk equivalents, which included either: (1) a predicted 10-year CVD risk ≥ 20% calculated using the PREVENT equation, or (2) the presence of very high-risk CKD according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Stage 3: Subclinical CVD. In the absence of direct biomarkers, this stage was operationally defined by CKM risk equivalents, which included either: (1) a predicted 10-year CVD risk ≥ 20% calculated using the PREVENT equation, or (2) the presence of very high-risk CKD according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Detailed staging criteria and the PREVENT equations are provided in Supplementary Tables S1 and S2, respectively. Covariate data collection Baseline data were collected via structured interviews, physical examinations, and laboratory tests. Covariates included sociodemographic characteristics, lifestyle factors, and anthropometric measurements like height, weight, waist circumference (WC), and systolic and diastolic blood pressure (SBP and DBP), from which body mass index (BMI) was calculated. Fasting blood was analyzed for fasting plasma glucose (FBG), glycated hemoglobin (HbA1c), C-reactive protein (CRP), hemoglobin (HGB), triglyceride (TG), low-density lipoprotein (LDL), high-density lipoprotein (HDL) and serum creatinine (Scr). Estimated glomerular filtration rate (eGFR) was calculated using the Asian Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. Prevalent conditions were ascertained, including hypertension (BP ≥ 130/80 mmHg or antihypertensive medication), diabetes (self-report, FBG ≥ 7.0 mmol/L, or HbA1c ≥ 6.5%), metabolic syndrome (MetS) (≥ 3 criteria), and CKD. Frailty was assessed using a validated cohort index, and relevant medication use was recorded. Air pollutant exposure assessment Daily air pollutant data were obtained from the high-resolution China High Air Pollutants (CHAP) dataset18–22. The CHAP dataset employs advanced multi-source satellite remote sensing and artificial intelligence techniques to capture the spatial and temporal heterogeneity of air pollution. By integrating extensive ground-based observational data, satellite remote sensing products, atmospheric reanalysis, and model simulations, the CHAP dataset provides high-resolution measurements of multiple air pollutants, including PM2.5, particulate matter with diameter ≤ 1 μm (PM1), particulate matter with diameter ≤ 10 μm (PM10), NO2, and O3. Participant baseline residences (2011) were geocoded and linked to the corresponding 1-km (for PM2.5, PM10, PM1, and O3) or 10-km (for NO2) grid cells. Long-term baseline exposure was defined as the mean pollutant concentration during 2010–2011, assuming residential stability during follow-up. Annual exposure levels from 2010 to 2017 were also assessed to characterize exposure trends. All concentrations were expressed in µg/m³. Outcome definition and ascertainment The outcome of interest was the incidence of CVD during the follow-up period (Waves 2–4, 2013–2018). At each wave, participants were asked standardized questions: “Has a doctor ever told you that you had heart problems (including heart failure, coronary heart disease, angina, heart attack, or other heart conditions)?” and “Has a doctor ever told you that you had a stroke?”. Participants who answered “yes” to either item during follow-up, but not at baseline, were classified as having incident CVD. The date of onset was estimated as the midpoint between the last follow-up without a diagnosis and the first follow-up with a reported diagnosis. Participants who remained free of CVD were censored at their last interview date. The validity of self-reported diagnoses in the CHARLS cohort has been established in prior studies23. Statistical analysis Baseline characteristics were summarized using means ± standard deviation (SD) or medians (interquartile range [IQR]) for continuous variables and frequencies (%) for categorical variables. We used Cox proportional hazards models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between air pollutants (modeled continuously and categorically) and incident CVD. The proportional hazards assumption was verified using Schoenfeld residuals. We constructed three progressively adjusted models: Model 1 (unadjusted), Model 2 (adjusted for demographic, socioeconomic, and lifestyle factors, including age, gender, urban/rural residence, geographic region, smoking statues, drinking statues, education level, occupation, marital status, frailty statues, cooking fuel use, BMI, SBP), and Model 3 (additionally adjusted for clinical and laboratory variables, including HGB, TG, LDL, HDL, HbA1c, CRP, eGFR, and CKM syndrome stage). Potential nonlinear relationships were explored using restricted cubic splines (RCS), with segmented regression used to identify thresholds where applicable. We also fitted time-varying Cox models using annually updated exposures; specifically, annual average concentrations for PM2.5 and NO2, and warm-season (April–September) averages for O3. To assess pollutant interactions and mixture effects, we constructed multipollutant models, included interaction terms, and applied Bayesian kernel machine regression (BKMR). BKMR was used to estimate the overall mixture effect of PM2.5, NO2, and O3 and determine individual pollutant importance via posterior inclusion probabilities (PIPs). Finally, we conducted subgroup analyses stratified by age, sex, region, smoking status and cooking fuel type, testing for effect modification with interaction terms. To evaluate the robustness of our findings, we performed several sensitivity analyses. These included: (1) applying 1- and 2-year exposure lags; (2) excluding participants who developed CVD within the first two years of follow-up; (3) restricting the analysis to rural participants; (4) employing a mixed-effects Cox proportional hazards model with a random intercept at the city level to control for spatial clustering; and (5) substituting PM10 and PM1 as alternative particulate matter exposure metrics. All statistical tests were two-sided, with a significance threshold of p < 0.05. All analyses and map visualizations were performed using R software (version 4.3.0). The administrative boundary data (1:1,000,000 scale) used for mapping were obtained from the National Fundamental Geographic Information Database (https://www.webmap.cn), provided by the National Geomatics Center of China. Results Baseline characteristics of the study population The study included 6,321 participants (mean age 58.9 years, 52.7% female), of whom 64.3% were in CKM stage 3 (Table 1). Over a median follow-up of 6.9 years, 896 (14.2%) developed incident CVD. Compared to those who remained CVD-free, individuals who developed CVD were older, more likely female, and presented with a significantly worse baseline cardiometabolic risk profile. This included a higher prevalence of comorbidities (frailty, MetS, hypertension, CKD), adverse clinical and laboratory markers (e.g., higher BMI, SBP, lipids, glucose, CRP; lower eGFR), and greater medication use (all P < 0.05). Table 1Baseline characteristics of study participants.VariableOveralln = 6321Non-CVDn = 5425CVDn = 896P-valueCKM, n (%)< 0.0010268 (4.24)245 (4.52)23 (2.57)1681 (10.77)612 (11.28)69 (7.70)21,307 (20.68)1,133 (20.88)174 (19.42)34,065 (64.31)3,435 (63.32)630 (70.31) Age (year) 58.87 (9.34)58.68 (9.40)60.02 (8.89)< 0.001Sex, n (%)0.001Female3,334 (52.74)2,816 (51.91)518 (57.81)Male2,987 (47.26)2,609 (48.09)378 (42.19)MetS, n (%)2,495 (39.47)2,046 (37.71)449 (50.11)< 0.001Diabetes, n (%)975 (15.42)802 (14.78)173 (19.31)< 0.001Hypertension, n (%)3,470 (54.90)2,874 (52.98)596 (66.52)< 0.001CKD, n (%)461 (7.29)381 (7.02)80 (8.93)0.042Marital_status, n (%)0.18Married and living with a spouse5,341 (84.50)4,575 (84.33)766 (85.49)Married but living without a spouse246 (3.89)221 (4.07)25 (2.79)Single, divorced, and windowed734 (11.61)629 (11.59)105 (11.72)Education_Status, n (%)0.75Elementary school or below4,458 (70.53)3,822 (70.45)636 (70.98)Middle school or above1,863 (29.47)1,603 (29.55)260 (29.02)Occupation, n (%)0.085Agricultural work3,967 (62.80)3,428 (63.22)539 (60.22)Non-agricultural work2,350 (37.20)1,994 (36.78)356 (39.78)Smoking_Status, n (%)0.25Non-smoker3,825 (60.52)3,267 (60.23)558 (62.28)Smoker2,495 (39.48)2,157 (39.77)338 (37.72)Drinking_Status, n (%)0.062Drink but less than once a month490 (8.19)429 (8.35)61 (7.18)Drink more than once a month1,331 (22.24)1,163 (22.64)168 (19.79)Non-drinker4,165 (69.58)3,545 (69.01)620 (73.03)Frailty, n (%)< 0.001Frailty639 (10.11)508 (9.36)131 (14.62)Robust5,682 (89.89)4,917 (90.64)765 (85.38)Residence, n (%)0.56Rural4,193 (66.33)3,591 (66.19)602 (67.19)Urban2,128 (33.67)1,834 (33.81)294 (32.81)Regional_Category, n (%)0.057East2,112 (33.41)1,833 (33.79)279 (31.14)Midland2,261 (35.77)1,909 (35.19)352 (39.29)West1,948 (30.82)1,683 (31.02)265 (29.58)Cooking_Fuel_Use, n (%)< 0.001Clean fuel1,974 (73.77)1,709 (74.96)265 (66.92)Solid fuel702 (26.23)571 (25.04)131 (33.08) SBP (mmHg) 128.75 (20.90)128.01 (20.60)133.20 (22.12)< 0.001 DBP (mmHg)) 74.92 (12.01)74.62 (11.86)76.75 (12.72)< 0.001 BMI (kg/m 2 ) 23.49 (9.53)23.37 (10.13)24.24 (4.40)< 0.001 WC (cm) 84.81 (9.87)84.37 (9.67)87.48 (10.60)< 0.001 HGB (g/L) 143.75 (22.43)143.66 (22.42)144.31 (22.52)0.63 TG (mmol/L) 1.45 (1.18)1.44 (1.18)1.54 (1.16)< 0.001 HDL (mmol/L) 1.34 (0.39)1.35 (0.39)1.30 (0.39)< 0.001 LDL (mmol/L) 3.03 (0.90)3.02 (0.90)3.10 (0.92)0.013 UA (umol/L) 263.45 (73.61)263.72 (73.30)261.77 (75.44)0.26 FBG (mmol/L) 6.06 (1.97)6.03 (1.92)6.24 (2.20)0.003HbA1c (%)5.27 (0.83)5.25 (0.80)5.38 (0.95)< 0.001 CRP (mg/L) 2.73 (7.73)2.71 (7.81)2.85 (7.20)< 0.001 Scr (umol/L) 68.93 (21.32)69.07 (22.07)68.12 (15.98)0.27 eGFR (ml/min/1.73 m²) 92.41 (14.58)92.57 (14.64)91.47 (14.25)0.009Hypertension_medication, n (%)< 0.001No5,325 (84.39)4,647 (85.82)678 (75.75)yes985 (15.61)768 (14.18)217 (24.25)Diabetes_medication, n (%)< 0.001No6,105 (96.71)5,261 (97.12)844 (94.20)yes208 (3.29)156 (2.88)52 (5.80)Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; MetS, Metabolic syndrome; CKD, Chronic kidney disease; SBP, Systolic blood pressure; DBP, Diastolic blood pressure; BMI, Body mass index; WC, Waist circumference; HGB, Hemoglobin; TG, Triglycerides; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; FBG, Fasting blood glucose; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; Scr, eGFR, Estimated glomerular filtration rate. Baseline characteristics of study participants. Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; MetS, Metabolic syndrome; CKD, Chronic kidney disease; SBP, Systolic blood pressure; DBP, Diastolic blood pressure; BMI, Body mass index; WC, Waist circumference; HGB, Hemoglobin; TG, Triglycerides; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; FBG, Fasting blood glucose; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; Scr, eGFR, Estimated glomerular filtration rate. Exposure distribution and temporal trends of air pollutants Baseline exposures to PM2.5 and NO2 were significantly higher among participants who later developed CVD compared to those who did not (both P < 0.05); no such difference was observed for O3 exposure (Table S3). Spatially, PM2.5 and NO2 concentrations were highest in eastern and central China, while summer O3 was elevated in northern and western regions. Over the study period, national ambient PM2.5 and NO2 levels declined, whereas O3 levels steadily increased (Fig. 2, Figure S1). Spatial distribution of air pollutants in China. Panels show the geographic variation in PM2.5 (A), NO2 (B), and warm-season O3 (C) concentrations in 2010. Pollution estimates were derived from high-resolution satellite-based models. Map generated using R software (v4.3.0) with administrative boundaries from the National Geomatics Center of China (1:1,000,000 database; https://www.webmap.cn). Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone. Associations between baseline air pollutants and incident CVD After full adjustment, each 10 µg/m3 increase in baseline PM2.5 and NO2 concentrations was associated with a significantly higher risk of incident CVD (PM2.5: HR = 1.20, 95% CI: 1.13–1.28; NO2: HR = 1.20, 95% CI: 1.07–1.35; both P < 0.05). In contrast, no significant association was observed for baseline O3 exposure (HR = 1.00, 95% CI: 0.76–1.31, P = 0.99). Results were consistent when exposures were analyzed by quartiles (Table 2). Table 2Hazard ratios for incident CVD by baseline air pollution exposure in CKM 0–3.Air PollutantModel 1Model 2Model 3HR (95%CI)P valueHR (95%CI)P valueHR (95%CI)P value baseline PM 2.5 Continuous1.11 (1.07, 1.16)< 0.0011.22 (1.14, 1.30)< 0.0011.20 (1.13, 1.28)< 0.001QuartileQ1ReferenceReferenceReferenceQuartileQ21.23 (1.00, 1.50)0.0461.10 (0.77, 1.56)0.6041.06 (0.74, 1.51)0.750QuartileQ31.08 (0.88, 1.33)0.4491.29 (0.93, 1.81)0.1311.26 (0.89, 1.77)0.192QuartileQ41.78 (1.47, 2.14)< 0.0012.55 (1.85, 3.52)< 0.0012.42 (1.74, 3.36)< 0.001QuartileP for trend< 0.001< 0.001< 0.001 baseline NO 2 Continuous1.14 (1.07, 1.22)< 0.0011.23 (1.10, 1.38)< 0.0011.20 (1.07, 1.35)0.002QuartileQ1ReferenceReferenceReferenceQuartileQ21.01 (0.83, 1.22)0.9591.07 (0.77, 1.49)0.6721.04 (0.74, 1.45)0.836QuartileQ30.97 (0.79, 1.18)0.7591.12 (0.81, 1.56)0.4921.08 (0.77, 1.51)0.661QuartileQ41.42 (1.19, 1.71)< 0.0011.68 (1.21, 2.33)0.0021.54 (1.11, 2.15)0.011QuartileP for trend< 0.0010.0020.010 baseline O 3 Continuous0.93 (0.83, 1.05)0.2631.01 (0.77, 1.31)0.9601.00 (0.76, 1.31)0.993QuartileQ1ReferenceReferenceReferenceQuartileQ21.04 (0.86, 1.26)0.6830.88 (0.62, 1.25)0.4740.88 (0.62, 1.25)0.466QuartileQ31.23 (1.03, 1.47)0.0261.21 (0.88, 1.67)0.2361.21 (0.87, 1.68)0.255QuartileQ40.96 (0.79, 1.17)0.6980.95 (0.63, 1.44)0.8230.95 (0.62, 1.43)0.794QuartileP for trend0.7860.4100.468Model 1: crude model.Model 2: adjusted for age, gender, residence, regional category, smoking statues, drinking statues, education level, occupation, marital status, frailty, cooking fuel use, BMI, SBP.Model 3: Model 2 + HGB, TG, LDL, HDL, HbA1c, CRP, eGFR, CKM stage.Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; BMI, Body mass index; HR, Hazard ratio; CI, Confidence interval; SBP, Systolic blood pressure; HGB, Hemoglobin; TG, Triglyceride; LDL, Low-density lipoprotein; HDL, High-density lipoprotein; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; eGFR, Estimated glomerular filtration rate. Hazard ratios for incident CVD by baseline air pollution exposure in CKM 0–3. Model 1: crude model. Model 2: adjusted for age, gender, residence, regional category, smoking statues, drinking statues, education level, occupation, marital status, frailty, cooking fuel use, BMI, SBP. Model 3: Model 2 + HGB, TG, LDL, HDL, HbA1c, CRP, eGFR, CKM stage. Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; BMI, Body mass index; HR, Hazard ratio; CI, Confidence interval; SBP, Systolic blood pressure; HGB, Hemoglobin; TG, Triglyceride; LDL, Low-density lipoprotein; HDL, High-density lipoprotein; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; eGFR, Estimated glomerular filtration rate. Dose–response relationships and threshold identification Exposure-response analyses revealed a nonlinear relationship for both PM2.5(P nonlinearity = 0.012) and NO2 (P nonlinearity = 0.043) with incident CVD. The risk of CVD increased markedly when pollutant concentrations exceeded estimated thresholds of 39.64 µg/m³ for PM2.5 and 39.86 µg/m³ for NO2, respectively. In contrast, the association for O3 showed a different pattern, with the highest risk observed at lower concentrations, although the confidence intervals were wide, indicating considerable uncertainty, particularly at the extremes of the exposure range (Figs. 3 and 4). Dose–response relationships for baseline air pollutants and incident CVD. Restricted cubic spline models show the dose–response relationships between baseline exposure to PM2.5 (A), NO2 (B) and O3 (C) and the risk of incident CVD. HRs are plotted along with 95% CIs across a range of pollutant concentrations. Models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, HDL, HbA1c, CRP, eGFR and CKM stage. Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; HR, hazard ratio; CI, confidence interval. Thresholds in the association of PM2.5 and NO2 with incident CVD. Segmented Cox proportional hazards models were used to evaluate the relationship between baseline PM2.5 and NO2 concentrations and incident CVD risk. Solid lines indicate the estimated HRs, with shaded areas representing the 95% CIs. Models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage. Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; HR, hazard ratio; CI, confidence interval. Subgroup analyses by demographics and region The adverse cardiovascular effects of PM2.5 and NO2 were consistent across subgroups defined by age, sex, smoking, and cooking fuel type (all P interaction > 0.05). However, we observed significant effect modification by geographic region, with stronger associations for both pollutants in eastern China compared to the western region (P interaction < 0.01) (Fig. 5). Subgroup associations between baseline air pollutants and incident CVD. HRs and 95% CIs for baseline exposure to PM2.5 (A) and NO2 (B) were estimated in predefined subgroups stratified by age, sex, geographic region, smoking status, and household cooking fuel. Models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage. *P < 0.05; **P < 0.01; ***P < 0.001. Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; HR, hazard ratio; CI, confidence interval. Associations based on time-varying exposure models In time-varying models using annually updated exposure metrics, all three pollutants were significantly associated with incident CVD. Notably, the association for O3 became statistically significant (HR = 1.20, 95% CI: 1.11–1.30), alongside robust associations for PM2.5 (HR = 1.20, 95% CI: 1.13–1.28) and NO2 (HR = 1.28, 95% CI: 1.14–1.44) (Table 3). Table 3Hazard ratios for incident CVD by time-varying exposure in CKM 0–3.Air PollutantModel 1Model 2Model 3HR (95%CI)P valueHR (95%CI)P valueHR (95%CI)P value time-varying PM 2.5 1.14 (1.10, 1.19)< 0.0011.22 (1.15, 1.29)< 0.0011.20 (1.13, 1.28)< 0.001 time-varying NO 2 1.20 (1.13, 1.29)< 0.0011.31 (1.17, 1.47)< 0.0011.28 (1.14, 1.44)< 0.001 time-varying O 3 1.12 (1.07, 1.17)< 0.0011.22 (1.12, 1.32)< 0.0011.20 (1.11, 1.30)< 0.001All hazard ratios are per 10-unit increase in pollutant concentration.Model 1: crude model.Model 2: adjusted for age, gender, residence, regional category, smoking statues, drinking statues, education level, occupation, marital status, frailty, cooking fuel use, BMI, SBP.Model 3: Model 2 + HGB, TG, LDL, HDL, HbA1c, CRP, eGFR, CKM stage.Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; BMI, Body mass index; HR, Hazard ratio; CI, Confidence interval; SBP, Systolic blood pressure; HGB, Hemoglobin; TG, Triglyceride; LDL, Low-density lipoprotein; HDL, High-density lipoprotein; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; eGFR, Estimated glomerular filtration rate. Hazard ratios for incident CVD by time-varying exposure in CKM 0–3. All hazard ratios are per 10-unit increase in pollutant concentration. Model 1: crude model. Model 2: adjusted for age, gender, residence, regional category, smoking statues, drinking statues, education level, occupation, marital status, frailty, cooking fuel use, BMI, SBP. Model 3: Model 2 + HGB, TG, LDL, HDL, HbA1c, CRP, eGFR, CKM stage. Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; BMI, Body mass index; HR, Hazard ratio; CI, Confidence interval; SBP, Systolic blood pressure; HGB, Hemoglobin; TG, Triglyceride; LDL, Low-density lipoprotein; HDL, High-density lipoprotein; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; eGFR, Estimated glomerular filtration rate. Joint exposure assessment: multipollutant Cox and BKMR models In multi-pollutant Cox models, the strong positive association between PM2.5 and incident CVD remained robust after adjusting for NO2 and/or O3 (Table 4). For example, in the PM2.5 + NO2 model, each 10 µg/m³ increase in PM2.5 was associated with a higher risk of CVD (HR = 1.36, 95% CI: 1.22–1.53). Conversely, likely due to the strong positive correlation between PM2.5 and NO2 (r = 0.82, Table S4), the coefficient for NO2 became negative when concurrently adjusted for PM2.5 (e.g., HR = 0.76, 95% CI: 0.63–0.93). To further explore combined pollutant effects, we examined pollutant interactions. Participants simultaneously exposed to high levels of PM2.5 and NO2 or O3 had the greatest risk of incident CVD, compared with those exposed to low levels of both pollutants (Fig. 6). These results indicate potential synergistic interactions between pollutants (P interaction < 0.05). Consistently, the BKMR further confirmed these findings, demonstrating a positive overall mixture effect on CVD risk and identifying PM2.5 as the dominant contributor (PIP = 1.00) (Figs. 7 and 8). Table 4Hazard ratios for incident CVD by multi-pollutant models in CKM 0–3.ModelExposureHR (95% CI)P valueP for Interaction PM 2.5 +NO 2 PM2.51.36 (1.22, 1.53)< 0.0010.025NO20.76 (0.63, 0.93)0.009 PM 2.5 +O 3 PM2.51.22 (1.14, 1.31)< 0.0010.001O30.77 (0.58, 1.04)0.085PM2.5+NO2 + O3PM2.51.36 (1.21, 1.53)< 0.001NO20.79 (0.64, 0.97)0.024O30.85 (0.62, 1.14)0.279All models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage.Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; BMI, Body mass index; HR, Hazard ratio; CI, Confidence interval; SBP, Systolic blood pressure; HGB, Hemoglobin; TG, Triglyceride; LDL, Low-density lipoprotein; HDL, High-density lipoprotein; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; eGFR, Estimated glomerular filtration rate. Hazard ratios for incident CVD by multi-pollutant models in CKM 0–3. All models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage. Abbreviations: CKM, Cardiovascular-Kidney-Metabolic; CVD, Cardiovascular disease; PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; BMI, Body mass index; HR, Hazard ratio; CI, Confidence interval; SBP, Systolic blood pressure; HGB, Hemoglobin; TG, Triglyceride; LDL, Low-density lipoprotein; HDL, High-density lipoprotein; HbA1c, Glycated hemoglobin; CRP, C-reactive protein; eGFR, Estimated glomerular filtration rate. Interactions between PM2.5 and NO2/O3 on CVD risk. Participants were categorized into exposure groups based on median splits of PM2.5 and NO2 (A) or PM2.5 and O3 (B). Cox proportional hazards models were used to estimate HRs and 95% CIs for each exposure combination. Models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage. *P < 0.05; **P < 0.01; ***P < 0.001. Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; HR, hazard ratio; CI, confidence interval. Combined effect of PM2.5, NO2, and O3 on Incident CVD. The figure shows the estimated overall CVD risk across the quantiles of the joint exposure to PM2.5, NO2, and O3 using Bayesian kernel machine regression (BKMR). Models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage. Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone. Relative importance of PM2.5, NO2, and O3 in predicting incident CVD. Posterior inclusion probabilities (PIPs) were calculated using Bayesian kernel machine regression to quantify the relative contribution of each pollutant to the overall mixture effect. Models were adjusted for age, sex, residence, regional category, smoking status, drinking status, education status, occupation, marital status, frailty, cooking fuel use, BMI, SBP, HGB, TG, LDL, HDL, HbA1c, CRP, eGFR and CKM stage. Abbreviations: PM2.5, fine particulate matter ≤ 2.5 μm; NO2, nitrogen dioxide; O3, ozone; PIP, posterior inclusion probability. Sensitivity and secondary analyses Our primary findings were robust across a series of pre-specified sensitivity analyses. The associations of PM2.5 and NO2 with incident CVD remained consistent when applying 1- and 2-year exposure lags and after excluding cases diagnosed within the first two years of follow-up (Tables S7-S10). Restricting the analysis to rural participants, a subgroup with expected lower residential mobility, yielded very similar effect estimates, supporting the stability of our findings against potential migration-related bias (Table S11). Furthermore, to address potential residual spatial confounding beyond the broad regional adjustment, we employed a mixed-effects Cox model with a random intercept at the city level. The results from this model confirmed that the associations remained stable, suggesting our findings were not substantially influenced by spatial clustering (Table S12). Consistent results were also observed when substituting PM2.5 with PM10 and PM1 as alternative exposure metrics (Table S13-14, Figures S2-10). Notably, PM1 exhibited a non-linear dose-response pattern with an apparent threshold similar to that of PM2.5(Figures S4-5). Finally, multipollutant and BKMR analyses confirmed these findings, highlighting PM10 and PM1 as dominant contributors to CVD risk within their respective pollutant mixtures (Figures S7-10). Discussion In this nationwide cohort of Chinese adults in CKM stages 0–3, long-term exposure to air pollutants (PM2.5, NO2, and O3) was significantly associated with increased incident CVD risk. PM2.5 emerged as the most robust predictor in multipollutant and mixture models, which confirmed a significant, dose-dependent effect of combined exposures. These findings establish individuals in early CKM stages as a key population vulnerable to the cardiovascular harms of air pollution, highlighting a critical opportunity for primary prevention. Although current CKM frameworks largely overlook environmental factors, our study underscores that air pollution is a key modifiable risk. Prior studies have linked pollutants to CKM-related mortality or focused on narrow subgroups like prehypertensive individuals15,24. Our findings fill a critical gap by demonstrating that ambient air pollution independently contributes to cardiovascular risk in early-stage CKM spectrum, warranting its inclusion in both clinical risk stratification and public health policy. Our finding that PM2.5 increases cardiovascular risk aligns with major international cohorts25–28, yet we extend this knowledge by identifying a nonlinear dose-response relationship with an apparent risk threshold near 40 µg/m3. While this value reflects a statistical inflection point rather than a true biological threshold, it is mechanistically plausible. High PM2.5 concentrations can overwhelm endogenous antioxidant and anti-inflammatory defenses, inducing chronic oxidative stress and systemic inflammation that promote endothelial dysfunction and atherosclerosis29–32. These processes could explain the steeper risk gradient observed at higher exposures. This finding has significant policy implications. Although the WHO’s 5 µg/m3 guideline remains the optimal target11, our empirically derived threshold—slightly above China’s current standard (35 µg/m3)—suggests that reducing PM2.5 below 40 µg/m3 could yield substantial near-term cardiovascular benefits. This offers an evidence-based interim goal to guide cost-effective air quality management in high-exposure regions. NO2, a traffic-related combustion pollutant and a recognized cardiovascular risk factor12, emerged as an important predictor of incident CVD in single-pollutant models. However, its independent effect is difficult to isolate because NO2 and PM2.5 are strongly correlated (r = 0.82; Table S4) due to shared emission sources. In multipollutant Cox models, this high degree of collinearity led to an unexpected negative coefficient for NO2—an implausible finding that likely reflects coefficient instability and differential exposure measurement error rather than a true protective effect. This interpretation was supported by collinearity diagnostics (variance inflation factor [VIF] values exceeding accepted thresholds; Table S5) and by a sensitivity model excluding NO2, in which the PM2.5 association remained unchanged (Table S6). Together, these results underscore the limitations of conventional regression methods in the presence of highly correlated exposures and highlight the value of mixture modeling approaches, such as BKMR, which identified PM2.5 as the dominant contributor while demonstrating a positive overall joint effect on CVD risk. While the cardiovascular effects of O3 are debated33–37, our findings offer important clarification. We observed a significant positive O3–CVD association only in the time-varying model, not the baseline model. This divergence likely reflects temporal exposure misclassification. During our study period (2010–2018), China experienced declining PM2.5 but rising O3 levels due to shifts in atmospheric photochemistry38, a trend evident in our cohort (Figure S1). Consequently, baseline O3 values systematically underestimated long-term exposure, whereas the time-varying model more accurately captured the cumulative burden, consistent with other large Chinese cohorts39. The U-shaped dose-response curve for O3 also requires cautious interpretation. This pattern may be an artifact of measurement error or confounding by co-pollutants, as low O3 levels can occur in high-traffic areas where it is titrated by nitric oxide (NO). Although a true hormetic effect cannot be excluded, supporting toxicological evidence is scarce. These findings underscore the importance of dynamic exposure modeling for O3 in rapidly evolving pollution environments like China. Our multipollutant models revealed significant synergistic interactions between PM2.5 and gaseous pollutants (NO2 or O3), leading to a greater-than-additive CVD risk. This highlights the importance of targeting the entire pollutant mixture. BKMR analysis further supported this, identifying PM2.5 as the dominant driver of a positive overall mixture effect. These results emphasize that public health interventions should address the combined impact of multiple pollutants, not just particulate matter alone. Our subgroup analyses revealed significant regional heterogeneity, with stronger pollutant-CVD associations in the more industrialized eastern regions, consistent with prior research40. In contrast, the lack of significant effect modification by age, sex, or smoking status highlights the pervasive nature of this risk across diverse demographic groups. These dual findings underscore the need for both broad and region-specific environmental health policies. Air pollution likely accelerates CKM progression through a stage-specific cascade of oxidative stress, inflammation, and endothelial–microvascular injury. In CKM stages 0–1, inhaled pollutants induce low-grade systemic inflammation and oxidative stress, promoting insulin resistance and initial microvascular impairment8,13,29. This metabolic vulnerability facilitates the transition to CKM stage 2, where established risk factors (e.g., hypertension, diabetes) amplify pollutant-induced injury. In this stage, excess reactive oxygen species (ROS) activate redox-sensitive pathways, such as nuclear factor-κB (NF-κB), elevating inflammatory markers like CRP, interleukin-6 (IL-6) and upregulating adhesion molecules, including vascular cell adhesion molecule-1 (VCAM-1)29–32,41,42. Concurrently, oxidative stress uncouples endothelial nitric oxide synthase (eNOS), reducing nitric oxide bioavailability and exacerbating microvascular dysfunction43,44. These synergistic insults drive arterial stiffness and subclinical atherosclerosis, accelerating the critical progression from metabolic risk (stage 2) to subclinical CVD (stage 3)30[,45. Our study benefits from the use of two high-quality national datasets (CHARLS and CHAP) and advanced statistical methods, such as time-varying and mixture models (BKMR), to assess dynamic and synergistic pollutant effects. We also applied a standardized CKM staging algorithm based on current AHA guidelines, offering policy-relevant evidence for a vulnerable, early-stage population in a high-exposure setting. Despite these strengths, several limitations should be noted. First, the cross-sectional effect sizes observed in our study were relatively small, which may limit clinical relevance. Second, CVD outcomes relied on self-reported diagnoses. While generally valid in CHARLS, this approach is susceptible to recall bias and under-ascertainment, particularly in socioeconomically disadvantaged regions (e.g., Western China) with limited healthcare access. Furthermore, the absence of clinical validation limits outcome specificity. This differential misclassification likely biases overall estimates toward the null and may contribute to the regional heterogeneity observed in our results. Third, we assumed residential stability during follow-up. However, prior evidence indicates minimal migration among older Chinese adults (< 1% inter-provincial mobility)⁴⁶, and our findings remained robust in the highly stable rural subgroup (Table S11), suggesting negligible migration bias46. Fourth, unmeasured co-pollutants—including volatile organic compounds (VOCs), persistent organic pollutants (POPs), heavy metals, and microplastics—often geographically co-cluster with PM2.5 due to shared emission sources and atmospheric transformation processes47,48. Thus, our estimates likely capture the cumulative toxicity of this broader environmental mixture rather than PM2.5 in isolation, highlighting the need for future research to employ high-resolution exposome assessments to disentangle these complex effects. Fifth, the absence of biomarkers like N-terminal pro-B-type natriuretic peptide (NT-proBNP) and troponins limited the precision of CKM staging. Finally, our findings in middle-aged and older Chinese adults may not be generalizable to younger populations or other ethnic groups. Our study identifies long-term ambient air pollution as a significant, modifiable risk factor for incident CVD in early-stage CKM syndrome. This underscores the urgent need to integrate environmental health into CKM management, using CKM staging as a framework for environmental-cardiometabolic risk stratification. By embedding pollution metrics in CKM risk models, healthcare systems can better address individual and environmental factors. Moreover, combining personal preventive measures (e.g., mask use and reduced outdoor activity during pollution peaks) with population-level strategies (e.g., urban planning, clean-fuel transitions, and air quality monitoring) may substantially mitigate exposures and alleviate CVD burden in CKM populations. Conclusions Our study establishes long-term ambient air pollution as a significant, modifiable risk factor for incident CVD in early-stage CKM syndrome, underscoring the imperative to integrate environmental health into CKM management frameworks. By leveraging CKM staging for environmental-cardiometabolic risk stratification, clinical and public health strategies can target vulnerable populations through multi-pronged interventions—including individual behaviors and systemic policies. Future research should focus on elucidating biological mechanisms and evaluating mitigation strategies to drive sustainable reductions in the global CVD burden. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We gratefully acknowledge the contributions of the entire China Health and Retirement Longitudinal Study (CHARLS) team, including the staff and participants who made this work possible. Abbreviations American Heart Association Body mass index Chronic kidney disease Chronic Kidney Disease Epidemiology Collaboration Cardiovascular-Kidney-Metabolic C-reactive protein Cardiovascular disease Diastolic blood pressure Estimated glomerular filtration rate Endothelial nitric oxide synthase Fasting blood glucose Glycated hemoglobin High-density lipoprotein Interleukin-6 Kidney Disease: Improving Global Outcomes Low-density lipoprotein Metabolic syndrome Nuclear factor-κB Nitric oxide Nitrogen dioxide N-terminal pro-B-type natriuretic peptide Ozone Particulate matter ≤ 1 μm Particulate matter ≤ 10 μm Fine particulate matter ≤ 2.5 μm Persistent organic pollutants Reactive oxygen species Systolic blood pressure Serum creatinine Triglycerides Uric acid Vascular cell adhesion molecule-1 Variance inflation factor Volatile organic compounds Waist circumference World Health Organization Author contributions All the authors solely contributed to this paper; WL and CX designed this study; WL wrote the manuscript; WL, FFL, XYL and WFY extracted and collated the data, and performed the analyses; YSL, ZJL, HSL, XFH and PYY acquisition of data; YZK and CX reviewed and edited this manuscript. Funding This research was supported by the Guangdong Health-Economics Medical Development Foundation Scientific Research Fund (Grant No. K-20240108). Data availability Data are available in the China Health and Retirement Longitudinal Study repository [http://charls.pku.edu.cn] and China High Air Pollutants datasets [https://weijing-rs.github.io/product.html]. Declarations Competing interests The authors declare no competing interests. Ethics approval and consent to participate The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Peking University (IRB00001052-11015). All participants provided written informed consent. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Data are available in the China Health and Retirement Longitudinal Study repository [http://charls.pku.edu.cn] and China High Air Pollutants datasets [https://weijing-rs.github.io/product.html].