Introduction

With the escalation in population aging, cardiometabolic multimorbidity (CM), regarded as the existence of at least two cardiometabolic diseases, is increasing in prevalence and is a major public health challenge1. Compared with patients with a single chronic disease, patients with CM have a lower quality of life, a greater susceptibility of death, and consume more healthcare resources, which creates many challenges for chronic disease prevention, control, and management2,3.

There are numerous risk factors for CM, such as behavioral risk factors, which include smoking, inadequate physical activity, and an unhealthy diet etc. Except for these established factors, exposure to outdoor air pollution also strengthens the susceptibility of CM morbidity and mortality4,5. Air pollution is thus a paramount global public health problem, especially as there is substantial evidence that air pollution is an intervenable unfavourable factor for cardiovascular disease (CVD) and ranks as the fourth most preventable modifiable risk factor contributing to the overall disease burden. Ambient air pollution usually includes particulate matter (PM) with diameters ranging from ≤ 2.5 μm to ≤ 10 μm (PM2.5-10), nitrogen dioxide (NO2), and nitrogen oxide (NO).

Previous studies have mostly focused on individual air pollutants; however, people are often exposed to a composite of pollutants, and their potential combined effects on the development of CM are largely unknown. One study examined long-term exposure to individual air pollutants (PM2.5 and NO2) and found it was related with an elevated risk of progressing from a disease-free state to having a single cardiometabolic disease and next to CM and death5. However, the effects of individual air pollutants may differ from their combined effect. Specifically, synergism between pollutants can result in their combined effect being greater than the sum of their individual effects6. To date, no studies have examined the possible influence of the combined exposure to air pollutants on the trajectory from a single cardiometabolic disease to CM and next to death. Besides, the association between air pollutants and CM appears to be influenced by genetic factors7. However, whether genetic factors influence the association between combined exposure to air pollutants and the development of CM remains unknown.

In the current study, we employed a multistate model to value the combined effect of exposure to various air pollutants on the trajectory of CM development, including the progression from a healthy state to the first instance of cardiometabolic disease, next to CM, and eventually to all-cause mortality. We also tested whether genetic susceptibility modified this effect. Our results contribute to epidemiological understanding of pollutant synergism that will aid the effective management of multi-pollutant air quality.

Methods

Study population

As a large biomedical database, the UK Biobank is applied to conduct detailed investigations of the lifestyle and genetic determinants of disease in middle-aged and older adults8. The study had approximately 500,000 individuals, who were recruited from 2006 to 20l0 from 22 centers in Scotland, England, and Wales, and 94% had European ancestry. The database includes genome-wide genotyping, clinical examination results, biological samples, and detailed baseline information on the individuals and is linked to individual health records. It also includes informed consent forms electronically signed by all of the individuals. The questionnaires and interviews collected the individuals’ sociodemographic information, family history and early-life exposure factors, psychosocial factors, lifestyle factors, and health status. This large prospective study was approved by the Northwest Research Ethics Committee (ref: 11/NW/0382). The present study used UK Biobank resources (application ID 78619).

Supplementary Fig. S1 presents the selection process. First, we excluded 3,272 of the 502,412 individuals who had type 1 diabetes or gestational diabetes; next, we excluded 44,862 individuals who did not have air pollutant exposure data. The remaining 458,118 eligible individuals who had complete data were included in our study.

Air pollution exposure assessment

Land-use regression (LUR) models developed by the European Study of Cohorts for Air Pollution Effects project were utilized to calculate the annual average air pollution exposures for individuals in the UK Biobank9. The residential addresses of all participants were geocoded and linked to air pollution data. The LUR techniques, established by the ESCAPE project, provided accurate estimates for particulates within a 400 km radius and required a spatial resolution of at least 100 m. These models incorporated geographic information system-derived predictors, such as traffic intensity, and were validated using monitoring data. Model accuracy was assessed using the leave-one-out cross-validation method, yielding cross-validation coefficient of determination (R2) values of 77% for PM2.5, 88% for PM10, 87% for NO2, and 88% for NO10.

Definition of the genetic risk score

The well-validated PRSs for 53 diseases were offered by the UK Biobank Polygenic Risk Score (PRS) Release11. We classified individuals into three groups in terms of their genetic risk of CVD, coronary heart disease (CHD), stroke, and type 2 diabetes (T2D): low-risk category, intermediate-risk category, or high-risk category.

Outcome assessment

The primary outcomes were CM, i.e., having two cardiometabolic diseases (T2D and CHD, T2D and stroke, or CHD and stroke) or all three diseases, and all-cause mortality. These outcomes were determined based on hospital inpatient records, self-reported information, and death registry records (Supplementary Table S1). The date of CM was the earliest date of occurrence of any of the three cardiometabolic diseases.

Air pollution score determination

The score = β1 × PM2.5 + β2 × PM2.5-10 + β3 × PM10 + β4 × NO2 + β5 × NO. The β coefficients were calculated using a multivariate-adjusted model with single air pollutants as independent variables. The score was from 45.58 to 174.11, with a upper score suggesting greater exposure to air pollution.

Covariates determination

At recruitment, information on individuals’ sociodemographic and lifestyle factors and their personal and family medical histories was collected via a self-administered questionnaire. The following covariates were selected for the models: age, sex, ethnicity (white/non-white), education (0–7 years, 8–10 years, 11–15 years, or ≥ 16 years), body mass index (BMI) (continuous), income (continuous), systolic blood pressure (SBP) (continuous), physical activity (metabolic equivalent minutes (MET-min)/week), work (employed/unemployed), alcohol consumption (never/past/current), and smoking status (never/past/current). Alcohol consumption and smoking status were self-reported. The MET-min/week was defined as the sum of all activity minutes per week. The Townsend deprivation index was used as a regional proxy indicator of socioeconomic status.

A healthy lifestyle score was constructed by seven variables: physical activity, BMI, alcohol consumption, smoking status, vegetable and fruit intake (servings/day), waist-to-hip ratio, and sedentary time (h/day). A detailed definition of the healthy lifestyle score has been previously elaborated12. Individuals were classified into three groups: an unhealthy low-score (0 or 1) group, an intermediate-score (2 or 3) group, and a high-score (≥ 4) group.

Statistical analysis

The correlations among single air pollutants were determined by Spearman’s correlation coefficients. Multivariate-adjusted Cox proportional hazards models were applied to estimate the relationship between combined exposure to air pollutants with CM and all-cause mortality. Model 1 was adjusted for age and sex; model 2 was additionally adjusted for ethnicity, education, income, work, alcohol consumption, and smoking status; model 3 was further adjusted for BMI, SBP, and physical activity.

We used a multistate regression model to calculate the HRs and 95% CIs for the association between air pollution score (Q1 vs. Q4) and the progression trajectories from being healthy to having one cardiometabolic disease and further to incident CM and all-cause mortality. This model allowed for the simultaneous estimation of the influence of risk factors on the progression from being healthy to having a single cardiometabolic disease and on the progression from having single cardiometabolic disease to CM13.

The lowest genetic risk group and the lowest air pollution quartile were used as the reference group for assessing the combined effect of the air pollutants and the GRS on the susceptibility of incident CM. Stratified analyses by age, sex, and healthy lifestyle score were also performed. Moreover, to estimate the robustness, we conducted the following sensitivity analyses. Thus, we (1) used an air pollution score derived from NO2, NO, PM10, and PM2.5; (2) eliminated incident CM cases occurring during the first 2 years of follow-up; (3) eliminated incident CM cases occurring during the first 5 years of follow-up; and (4) further modified the model for total cholesterol concentration.

All analyses were calculated by R software (version 4.0.3). All P-values lower than 0.05 were regarded as significance with two-side.

Results

Descriptive results

This study contained a sum of 458,118 individuals at baseline. After an average follow-up of 12.1 years, 5,178 (10.76%) individuals developed CM, and 1,238 (23.91%) died. Table 1 presents the baseline information of the individuals.

Table 1 Baseline characteristics of participants in the UK Biobank study.
Full size table

Association between the air pollution score and CM and all-cause mortality

The baseline characteristics in terms of air pollution score quintile is presented in Supplementary Table S2. The Spearman correlation coefficients among the air pollutants are indicated in Supplementary Table S3. The relationships between individual air pollutants and the risk of CM are displayed in Supplementary Table S4, which revealed that NO2 (p = 0.002), NO (p = 0.002) and PM2.5 (p = 0.018) were significantly associated with an increased risk of CM. The associations between the air pollution score and the risk of CM and all-cause mortality are displayed in Tables 2 and 3, respectively. After multivariable adjustment (i.e., in model 3), we observed that the air pollution score was associated with a greater susceptibility of CM and all-cause mortality in a dose-dependent style for individuals without cardiometabolic disease at baseline (p < 0.001) and for individuals with T2D (p < 0.001) or stroke (p = 0.003) at baseline.

Table 2 Adjusted HRs and 95% CIs for air pollution score with the risk of cardiometabolic multimorbidity.
Full size table
Table 3 Adjusted HRs and 95% CIs for air pollution score with the risk of all-cause mortality.
Full size table

Progression from no cardiometabolic disease at baseline to CM

The associations for the air pollution score and the susceptibility of progressions in CM development are presented in Fig. 1. Among the 411,063 healthy individuals at baseline, 48,120 (11.71%) developed one cardiometabolic disease. After multivariable adjustment (i.e., in model 3), the air pollution score (Q1 vs. Q4) was related with the susceptibility of progressing from a healthy state at baseline to having one cardiometabolic disease. It was also related with the susceptibility of subsequently progressing to CM and next to all-cause mortality. However, the air pollution score was not associated with the progression from CM to all-cause mortality. As shown in Fig. 1, the multivariable-adjusted HRs (95% CIs) of the air pollution score were 1.04 (1.03–1.05) for progression A, 1.04 (1.00–1.07) for progression B, 1.05 (1.03–1.06) for progression C, 1.07 (1.04–1.10) for progression D, and 0.95 (0.89–1.01) for progression E.

Fig. 1
figure 1

Relationships of air pollution score with susceptibility of all progressions from disease-free baseline healthy to one cardiometabolic disease, next to cardiometabolic multimorbidity and eventually to death.

Full size image

Progression from one cardiometabolic disease at baseline to CM

The relationships between the air pollution score and the susceptibility of progressing from having one cardiometabolic disease to CM are presented in Fig. 2. Among the 48,120 individuals with one cardiometabolic disease at baseline, 2,378 (15.93%) progressed from T2D to CM; 1,918 (7.42%) progressed from CHD to CM; and 882 (12.01%) progressed from stroke to CM. For the progression from CHD to CM and compared with the Q4 air pollution score, the adjusted HR (95% CI) of the Q1 air pollution score for CM (model 3) was 1.05 (1.00–1.10). The progressions from T2D or stroke to CM were not significantly associated with the air pollution score.

Fig. 2
figure 2

Relationships of air pollution score with susceptibility of all progressions from disease-free baseline healthy to one of specific cardiometabolic disease, next to cardiometabolic multimorbidity and eventually to death.

Full size image

Progression from no or one cardiometabolic disease at baseline to all-cause mortality

The HR (95% CI) of the air pollution score (Q1 vs. Q4) was 1.05 (1.03–1.06) for the progression from being healthy to all-cause mortality, and they were 1.06 (1.00–1.12), 1.09 (1.05–1.14), and 1.01 (0.95–1.06) for the progression from T2D, CHD, and stroke to all-cause mortality, respectively (Fig. 2). The progression from CM to all-cause mortality was not significantly associated with the air pollution score.

Combined influence of air pollutants and genetic susceptibility on the risk of CM and all-cause mortality

Additionally, we calculated the combined influence of the air pollution score and the CVD, CHD, stroke, and T2D GRS in terms of risk of CM development (Supplementary Table S5) and all-cause mortality (Supplementary Table S6). Individuals without CM and with a high GRS and high air pollution score had a greater susceptibility of CM and all-cause mortality than individuals without CM and with a low GRS and low air pollution score, even though there were no statistically significant interactions.

Subgroup and sensitivity analyses

After stratified by subgroups, the results showed that in individuals without CM, a greater air pollution score was related with a greater susceptibility of CM in women and with a lower healthy lifestyle score (Supplementary Table S7). However, in men without CM, a greater air pollution score was connected with a greater risk of all-cause mortality (p < 0.001) (Supplementary Table S8).

The results did not alter distinguishably when an air pollution score obtained from NO2, NO, PM10, and PM2.5 was used. Moreover, the results remained robust after the exclusion of incident cases of CM during the first 2 years (Supplementary Table S9) or first 5 years (Supplementary Table S10) of follow-up. Additional adjustment of model 3 for total cholesterol concentration also did not alter the results.

Discussion

Our findings indicate that after multivariable adjustment (i.e. in model 3), a high air pollution score was related with CM and all-cause mortality susceptibility in a dose-dependent style in individuals without any cardiometabolic disease at baseline and in those with T2D or stroke at baseline. Besides, the air pollution score (Q1 vs. Q4) was related with the risk of progressing from baseline healthy to having one cardiometabolic disease, from having one cardiometabolic disease to CM, and from CM to all-cause mortality. Additionally, we found that the associations for the air pollution score and the susceptibility of CM and all-cause mortality were strengthened by genetic susceptibility to CM; however, these associations were not statistically significant.

Compared to previous studies, the present research not only analyzes the direct relationship between air pollution indices and the risk of cardiovascular diseases but also delves deeply into the role of air pollution in the progression from baseline health status to the development of cardiometabolic diseases, and subsequently to cardiovascular diseases and all-cause mortality. This aligns with findings from prior research, which associate air pollution with various adverse health outcomes5,14. In addition, our study expands these findings by incorporating multiple air pollutants (PM2.5, PM10, NO2, and NO) and considering their combined effects. This is reflected by the World Health Organization Global Air Quality Guidelines, which state that research must be based on multi-pollutant approaches15.

Air pollutants not only have a part in the pathogenesis of cardiometabolic diseases but also in their susceptibility to CM and next to death. These pollutants may affect CM by triggering the inflammatory response or causing oxidative stress, vascular endothelial dysfunction, blood hypercoagulability and thrombosis, elevated blood pressure, atherosclerosis and cardiac remodeling, autonomic regulatory dysfunction, cardiac electrophysiological changes and arrhythmias, or metabolic syndrome and insulin resistance16. Although a meta-analysis indicates a positive correlation between air pollution and the combined outcome of stroke hospitalization and mortality in large-scale studies17, our study found no significant association between air pollution index and the progression from stroke to cardiovascular diseases or mortality. This may be related to stroke patients reducing their outdoor activities, thereby decreasing their exposure to harmful outdoor air pollutants18,19.

Regarding the progression from a specific cardiometabolic disease to CM, the progression from stroke to CM and the progression from stroke to death were not significantly associated with the air pollution score. However, a large meta‑analysis suggested that air pollution was positively related with a combined outcome of stroke hospital admission and mortality17. Stroke is characterized by a high rate of disability; however, improved medical care, continual advances in clinical diagnosis and treatment techniques, and the dissemination of stroke prevention and treatment knowledge has reduced stroke mortality18. This is partly attributable to stroke patients reducing their participation in outdoor activities, as this reduces their level of exposure to health-degrading outdoor air pollution19. Nevertheless, our results show that the progression from CHD to CM was significantly associated with the air pollution score, indicating that patients with CHD should be concerned about exposure to air pollution increasing their susceptibility of CM.

The stratified analysis showed that female hormones (especially estrogen) have a cardiovascular protective effect20. In addition, the internalized coping style that is sometimes adopted by women in response to stress makes women more susceptible than men to chronic diseases, such as cardiovascular disease21,22. Thus, sex differences in CM risk should be noted in clinical practice, as this will improve treatment and secondary prevention strategies. In men, a greater air pollution score was related with a superior susceptibility of all-cause mortality. Differences in estrogen concentrations are related to the observed differences in mortality between men and women, with women having a lower risk of mortality than men due to the protective effect of estrogen23. In addition, compared with women, men have a greater susceptibility of mortality due to a longer cumulative exposure to air pollution in their work environment.

Individuals with a lower healthy lifestyle score had a greater susceptibility of CM, suggesting that a healthy lifestyle may partly offset the harmful effects of air pollution. This is in keeping with previous studies, which found that PM2.5, PM10, and NO2 were jointly related with an greater susceptibility of cardiometabolic disease, especially in those with an unhealthy lifestyle, and that following a healthy lifestyle may extenuate the adverse effect of household use of polluting fuels on CM risk24. Taken together, these findings suggest that healthy lifestyles should be promoted as a means to lessen the harmful effects of air pollution.

In our study, individuals with a greater GRS and a greater air pollution score had a greater susceptibility of CM and all-cause mortality than individuals with a lesser GRS and lesser air pollution score, even though there were no statistically significant interactions. This may be attributable to the fact that the genetic variants that constituted the GRS in our study only accounted for a small part of the CM risk, and that genetics accounts for only a small proportion of the explainable risk of CM development. Thus, the air pollution score may be positively related with CM susceptibility regardless of genetics.

Interventions to control risk factors are important tools for solving the serious problems posed by CM, and effective preventive strategies should be implemented. Considering the different and complex sources, physical properties, and chemical compositions of air pollutants, there is a need for the establishment of a worldwide air pollution surveillance network and the strengthening of epidemiological research on cardiovascular damage caused by multiple pollutants and the mechanisms by which they generate this damage. The current guidelines recommend activity modifications, the use of indoor air purifiers, and the use of personal respiratory protection to reduce the harmful effects of air pollution exposure24.

This study highlights the significant impact of air pollution on cardiovascular mortality and overall mortality. The findings offer a foundation for further research on the link between cardiovascular mortality and exposure to air pollution. Additionally, they can inform policymakers in making evidence-based decisions. For the general public, this research sheds light on preventive measures against diseases. Thus, this study holds great importance.

A major strength of our study is that it used multistate models and air pollution scores based on multiple pollutants to comprehensively assess the influence of air pollution on the progression of CM and all-cause mortality. We obtained these air pollution scores by calculating Cox regression coefficients after modifying for all possible confounding factors. Thus, this study considered the interactions between five air pollutants and generated a multi-pollutant model that can estimate the influence of air pollution exposure on health more accurately than can single-pollutant models. The significance of assessing the health influences of multiple pollutant exposure has been progressively realized, as air pollutants may be extremely correlated and come from the same emission source25,26. Thus, compared with a single pollutant score, an air pollution score is a better confined quantification of exposure to air pollution, as it assesses the adverse effects on health due to concomitant exposure to multiple air pollutants27. In addition, the large sample size and the large number of covariates allowed us to clearly identify important influencing factors and to perform stratified analyses to investigate effect modifications. Furthermore, our results remained robust after a sensitivity analysis that adjusted for confounders.

The present study also has some limitations. First, similar to previous cross-sectional studies, we were unable to determine causal relationships between air pollution scores with CM and all-cause mortality. Second, our analysis was based on UK Biobank individuals aged 40–69 years, which may limit the extrapolation of our findings to groups comprising different ethnicities or age groups. Third, we could only estimate the individuals’ level of air pollutant exposure based on their residential address, as we did not have access to portable monitoring devices or indoor air-pollutant detection instruments to collect the individuals’ personal air-pollutant exposure data. Therefore, we used a LUR-based model to estimate the annual average concentration of air pollutants to which the individuals were exposed, which inevitably resulted in exposure measurement errors. These may have led to random errors that masked the true effect of the air pollutants, i.e., non-differential errors that masked the relationships between the air pollutants and CM and all-cause mortality. Finally, we corrected for a large number of covariates, some of which may have been mediating variables rather than confounders or nonessential confounders; thus, the model may have been over-corrected or unnecessarily corrected.

In conclusion, air pollutants were together related with a greater susceptibility of CM and all-cause mortality. We also found that air pollutant exposure was associated with an elevated susceptibility of progressing from a disease-free state to having a single cardiometabolic disease and with subsequent progressions to CM and death. These findings show that air pollution is an important and intervenable unfavourable factor for CM and all-cause mortality, thereby highlighting the need to decrease people’s exposure to air pollution.