Introduction

Obesity is a chronic disease that seriously harms health and is closely related to the occurrence of more than 50 diseases such as dyslipidemia, diabetes, hypertension, non-alcoholic fatty liver, and cardiovascular disease1,2,3. According to the data from The Global Burden of Disease (GBD) database, the age-standardized prevalence of overweight and obesity increased significantly from 26.5% to 7% in 1985 to 39% and 12.5% in 2015, respectively4. In China, a national survey showed that the prevalence of obesity increased from 3.1% in 2004 to 8.1% in 2018, with a steady increase among rural women5.

The risk of obesity attributable to socioeconomic status (SES) inequalities has long been reported6,7,8. SES can reflect a person’s position in the social economy, which is generally composed of indicators such as education level, personal income, occupation, and marital status9,10,11,12. It has been pointed out that low socioeconomic status may produce negative psychological emotions, which in turn may lead to emotional eating, which in turn can lead to obesity13. Some studies have shown an inverse association between SES and obesity prevalence in developed countries and a positive association between SES and obesity prevalence in developing countries14,15. In recent years, a four-stage model of global obesity development has been proposed. In the first stage, the prevalence of obesity was higher in women than in men, higher in people with higher socioeconomic status than in people with lower socioeconomic status, and higher in adults than in children. In the second phase of countries, the prevalence of obesity has increased significantly in adults, and small in children, and gender disparities and socioeconomic disparities among women are narrowing; In the third phase, the prevalence of obesity in people with lower SES was higher than that in people with higher SES. In the fourth stage, the prevalence of obesity decreases16.

Recently, more and more studies have shown that air pollutants are related to the occurrence of obesity17,18,19, including indoor and outdoor air pollutants20,21, which may increase the prevalence of obesity. Further studies showed that the effect of air pollutants on obesity was influenced by gender, age, and type of air pollutants22. The potential mechanisms by which air pollutants increase the risk of obesity may include inflammation, oxidative stress, metabolic imbalance, intestinal flora disorder, and epigenetic modification20, in addition, air pollutants may also induce sedentary behavior, chronic psychological stress and the activation of hypothalamic-pituitary-adrenal (HPA) axis, which may also increase the risk of obesity23. However, people with different SES often have different exposure levels of air pollutants in their living and working environments24,25.

However, the impact of SES on the relationship between air pollution exposure and obesity risk remains poorly understood. Thus, this study aimed to investigate the role of SES, encompassing household per capita monthly income, education level, and marital status, in the effect of PM2.5 and its components (SO42− : sulphate; NO3− : nitrate, NH4+ : ammonium, OM: Organic Matter; BC: Biological Component) exposure on obesity potential variable (BMI).

SO42−, NO3−, and NH4+ are key secondary aerosol components, closely related to human activities. Monitoring OM helps identify the relative importance of various organic pollution sources, and BC surveillance aids in understanding combustion’s contribution and source proportions in air pollution. Incorporating these into PM2.5 monitoring and analysis offers a more comprehensive understanding of atmospheric pollution composition.

From clinical and public health perspectives, clarifying these relationships is vital. Understanding how SES modulates the air pollution-obesity link can guide interventions. For instance, in areas with higher SES and strong air pollution-obesity links, personalized health education programs could target environmental and lifestyle factors for obesity. This knowledge can also support policies to reduce air pollution and address obesity disparities among different SES groups.

Materials and methods

Study population

Jinchang Cohort was used as a platform for this study. From June 2011 to December 2013, 48,001 subjects completed the baseline survey and the first follow-up was completed from January 2014 to December 2015. The average follow-up period was 2.3 years. Detailed information about the Jinchang Cohort has been presented previously26,27. After excluding 6017 subjects with obesity at baseline survey, 22,810 subjects who were lost to follow-up and had incomplete home address information, 19,174 subjects were finally included in the study. This study was approved by the Ethics Committee of Lanzhou University School of Public Health (batch number: 2017-01). All subjects signed the informed consent.

Measurement of air pollution exposure

Daily surface monitoring data of PM2.5, SO42−, NO3−, NH4+, OM, and BC from four environmental monitoring stations (Xinchuanyuan, Ministry of Transport, Municipal Science and Technology Commission, the Second Guest House of the Jinchuan Company) near the home address of the study population were collected by Jinchang Environmental Monitoring Station from January 1, 2011, to December 31, 2015. Missing data were imputed using the mean value of neighboring points of environmental monitoring stations. The time of health examination was used as the matching variable to match the average value of air pollutants data from the baseline examination time to the first follow-up examination time.

Individual pollutant exposure levels were estimated using the nearest neighbor model28. The home address of each subject was used as the matching variable to match the nearest station to the four environmental monitoring stations in Jinchang City, and the air pollutants level measured by the monitoring point was used to assess the individual exposure level. Google Map software was used to query the location of each subject’s residence and the latitude and longitude of the four monitoring sites. ArcGIS10.3 software was used to calculate the distance between each subject’s residence and the four monitoring sites. The monitoring site closest to the study subject was selected, and then environmental monitoring data were matched sequentially. If the relevant measurement data were not available at the nearest monitoring site, testing was performed from the second closest monitoring site until the exact monitoring data were available to assess individual air pollutant exposure.

Field investigation

Field investigation included a questionnaire survey, physical examination, and clinical biochemical index detection.

The questionnaire was self-designed and structured. After obtaining the informed consent of the subjects, the trained investigators conducted face-to-face interviews. The main information included general socio-demographic characteristics (gender, age, occupation, marital status, education level, family income, etc.), lifestyle habits (smoking, drinking, physical exercise, etc.), history of previous diseases (history of cardiovascular and cerebrovascular diseases, cancer, endocrine and metabolic diseases), and family history of diseases.

Physical examination and clinical biochemical indexes were performed by professionals of Jinchuan Company Staff Hospital. Physical examination included measurement of height, weight, and blood pressure. Height and weight were measured using a computerized body scale (SK-X80/TCS-160D-W/H) in a standard position without shoes. Blood pressure was measured using an electronic sphygmograph (BP705) produced by AMPall company in Korea. The subjects had to rest for more than 30 min before measurement, and the average of three measurements was taken as the final blood pressure value. The clinical biochemical indicators mainly included fasting blood glucose, total cholesterol, triglyceride, serum low-density lipoprotein cholesterol, serum high-density lipoprotein cholesterol, etc. The test instrument was an automatic biochemical analyzer (7600-020) produced by Hitachi. Among them, fasting blood glucose was measured by the oxidase method, and triglyceride was measured by the glycerol phosphate oxidase-peroxidase method.

Diagnostic criteria and related definitions

According to the health standard WS/T 428–2013 “Adult weight Determination” issued by the National Health Commission of the People’s Republic of China29, overweight was defined as 24.0 ≤ BMI<28.0 kg/m2, and obesity was defined as BMI ≥ 28 kg/m2.

Smoking was defined as ≥ 1 cigarette/day, continuous smoking>6 months. Alcohol consumption was defined as an average of ≥ 1 occasion/week, continuous drinking>6 months.

Socioeconomic status (SES) was composed of three dimensions: education level, family monthly income per capita, and marital status, which were assigned as follows: ① Education level: Junior middle school or below = 1, senior high school = 2, college or above = 3; ② Family per capita monthly income (yuan): <2000 = 1, 2000–4999 = 2, ≥ 5000 = 3; ③ Marital status: single = 1, Non-single = 2. SES scores range from 3 to 8, with higher scores indicating higher socioeconomic status9,10,11,12. The 33.3 and 66.6 percentiles of SES were selected as the cut-off points to divide SES into low, medium, and high levels.

Statistic analysis

Ambient air pollutant concentrations were divided into four levels according to their quartiles. The Time-Dependent Cox Regression Model was used to longitudinally analyze the PM2.5 and its component quartiles and their interquartile range(IQR) by 1% point increased Hazard Ratio (HR) and 95% confidence interval (CI) of obesity, respectively. Restricted cubic spline (RCS) plots were used to reflect the dose-response relationship between PM2.5 and its components and the risk of developing obesity. Time-dependent Cox regression models were used to estimate the HR and 95%CI of PM2.5 and its component quartiles with obesity in different levels of SES. Only one ambient air pollutant was included at a time, taking into account the collinearity between different air pollutants. Gender and age were first adjusted as model 1. Model 2 is fully adjusted. The basis of model 1, included ethnicity, occupation, physical exercise, smoking, drinking and tea, consumption of salted and dried foods, smoked foods, livestock meat, coarse grains, milk and dairy products, eggs, garlic, beans, and soy products, salad or fresh vegetables, fresh vegetables and fruits, diabetes mellitus, coronary heart disease, mental state in recent years, fasting blood glucose, total cholesterol, and glycerin Triglyceride, serum low-density lipoprotein cholesterol, serum high-density lipoprotein cholesterol, blood pressure.

Generalized Additive Models (GAM) were used to explore the interaction between PM2.5, its components, and SES on the risk of overweight/obesity. A three-dimensional (3D) map under the bivariate response surface model after adjusting for confounding factors was constructed and the interaction was analyzed qualitatively. To better assess the interaction between PM2.5, its components, and SES, this study divided PM2.5 and its components into four groups according to quartiles: Q1, Q2, Q3, and Q4, with Q1 and Q2 as low concentration groups and Q3 and Q4 as high concentration groups. The groups of PM2.5 and its components and SES levels were combined in pairs to form SES-air pollutants (low SES-Q1, medium SES-Q2, high SES-Q2, medium SES-Q3, high SES-Q3, medium SES-Q4, high SES-Q4) were included in the model to analyze the interaction between PM2.5, its components, and SES. The magnitude and direction of interaction were estimated using relative excess risk due to interaction (RERI), with low SES-Q1 as the reference. RERI>0 indicates a synergistic effect between PM2.5, its components, and SES, while RERI<0 indicates an antagonistic effect between PM2.5, its components, and SES30,31. The calculations are as follows:

$$\text{RERI = RR11 - RR10 - RR01 + RR00}$$

Where: RR11 is the RR value when the two exposure factors are present; RR01 and RR10 tables show the RR values when only one exposure factor was present; RR00 is the RR value when neither of the two exposure factors is present and is set to 1.

SPSS 22.0 and R 3.6.2 software were used for data analysis. A two-sided test was used, and the significance level was α = 0.05.

Result

Characteristics of the study population

A total of 19,174 subjects were included in the baseline study, including 10,742 males (56.0%) and 8432 females (44.0%), with an average age of 48.5 ± 8.3 years old. People with overweight or obesity were older than those without. People who are male in gender, have low education, work as a servicer or technician, smoke, drink, and drink tea are more likely to be overweight or obese (all P<0.05). The baseline characteristics of the subjects are shown in Table 1.

We conducted a gender-specific group test. As presented in Supplementary Table S1, a notable pattern emerges: among non-single individuals, male (57.58%) are in the majority. In higher-income cohorts, female (52.36%) predominate. The BMI of male (24.19 ± 2.50) is found to be higher than that of female (22.92 ± 2.64), and the general level of pollutant exposure in males also exceeds that of females(all P<0.05).

Table 1 Baseline characteristics of the study subjects [n/(%)/(\(\:\stackrel{-}{x}\)±S)].
Full size table

Distribution characteristics of PM2.5 and its components

Supplementary Fig S1 shows the descriptive statistics of pollutant concentration from the baseline survey to the first follow-up for each subject, and Fig. 1 shows the pair-to-pair correlation between PM2.5 and its components. The exposure levels of PM2.5, SO42−, NO3−, NH4+, OM, and BC during this period are respectively: 20.70–50.50 µg/m3, 2.73–6.39 µg/m3, 1.91–10.21 µg/m3, 1.81–6.03 µg/m3, 4.32–13.60 µg/m3, 0.98–2.13 µg/m3, the average exposure levels are: 37.04 µg/m3, 4.38 µg/m3, 4.87 µg/m3, 3.51 µg/m3, 7.14 µg/m3, 1.49 µg/m3. PM2.5 and its components were positively correlated, and the Spearman correlation coefficient (rs) was 0.65–0.98.

Fig. 1
figure 1

Spearman correlation between PM2.5 and its components (SO42-, NO3-, NH4+, OM, and BC ).

Full size image

Association of PM2.5 and its components with overweight/obesity

Table 2 shows the associations between PM2.5 and its components and the risk of overweight/obesity. After adjusting for demographic characteristics, lifestyle, dietary habits, and disease history the HR with 95%CI for overweight/obesity associated with exposure to the highest quartile of PM2.5 and its components, compared with the lowest quartile, were: 1.777 (1.406–2.246), 3.317 (2.626–4.190), 5.396 (4.180–6.964), 3.773 (2.950–4.825), 3.199 (2.497–4.098) and 3.371(2.655–4.281) (all P<0.05). For each interquartile range (IQR) increase in the concentration of PM2.5 and its components, the fully adjusted HR and 95% CI for being overweight/obese were: 1.229 (1.137–1.328), 1.512 (1.401–1.633), 1.814 (1.658–1.985), 1.545 (1.420–1.681), 1.492 (1.375–1.619) and 1.455 (1.347–1.570) (all P<0.05).

Table 2 HR (95% CI) for different concentrations of PM2.5 and its components (SO42-, NO3-, NH4+, OM, and BC ) and overweight/obesity after multivariable adjustment.
Full size table

As shown in Fig. 2, Restricted cubic spline analysis showed a non-linear dose-response relationship between exposure to PM2.5 and its components and the risk of overweight/obesity after controlling for the confounding factors included in model 2 (P for trend<0.05, P for non-linear<0.05 for all air pollutants). When NO3− concentration is> 5.14 µg/m3, NH4+ concentration is> 3.69 µg/m3, and OM concentration is> 7.61 µg/m3, the risk of developing overweight/obesity will show a significant upward trend with the increase of the concentration of these pollutants. PM2.5 concentration <36.37 µg/m3 slowed down the trend of increased risk of developing overweight/obesity as its concentration increased. When SO42− concentration <4.30 µg/m3, the tendency to increase the risk of overweight/obesity decreases with the increase of its concentration, while the tendency to increase the risk of overweight/obesity increases with the increase of its concentration when SO42− concentration >4.80 µg/m3. When BC concentrations>1.71 µg/m3, the risk of developing overweight/obesity showed an increasing trend as its concentration increased.

Fig. 2
figure 2

Association of PM2.5 and its components (SO4 2- , NO3 - , NH4 + , OM, and BC ) with overweight/obesity (The adjustment factors were the same as the model 2).

Full size image
Table 3 HR (95% CI) of PM2.5 and its components (SO42-, NO3-) and overweight/obesity were calculated with SES as the stratification variable.
Full size table
Table 4 HR (95% CI) of PM2.5’s components (NH4+, OM, BC ) and overweight/obesity were calculated with SES as the stratification variable (continued).
Full size table

Association of PM2.5 and its components with overweight/obesity after adjusting for SES

Table 3 shows the associations between PM2.5 and its components with the risk of overweight/obesity after adjusting for SES levels. After adjusting for demographic characteristics, lifestyle, dietary habits, and disease history, it was found that the risk of overweight/obesity increased with the increase of PM2.5 and its component concentration in different SES levels, and this effect was most pronounced at high SES. For example, in medium levels of SES, the HR with 95%CI of SO42−, Q2, Q3, Q4 corresponding to the risk of overweight/obesity compared with Q1 concentration are: 1.492 (1.229–1.812), 1.937 (1.397–2.686), 2.243 (1.394–3.610) (all P<0.05). At high levels of SES, the HR with 95%CI of SO42−, Q2, Q3, Q4 corresponding to the risk of overweight/obesity compared with Q1 concentration were: 1.855 (1.444–2.383), 4.023 (2.631–6.151), 7.446 (4.053–13.679) (all P<0.05). The HR and 95% CI of overweight/obesity were 1.338(1.207–1.484), 1.311(1.121–1.533), and 2.224(1.823–2.714) for every quartile range (IQR) increase in SO42− concentration at low, middle and high SES levels (all P<0.05).

Interaction between SES and air pollutants on overweight/obesity

Figure 3 shows the qualitative analysis results of the interaction between SES and PM2.5 and their components. At low SES, with the increase of PM2.5, SO42−, OM concentration, the risk of obesity presents a “U” pattern, which decreases first and then increases; with the increase of NO3− concentration, the risk of obesity increases, while with the increase of BC concentration, the risk of obesity decreases. At high SES, with the increase of PM2.5, SO42− and BC concentrations, the risk of obesity increases; with the increase of NO3− and OM concentrations, the risk of obesity increases first and then decreases in an inverted “U” shape. The risk of obesity rose as NH4 concentrations increased, regardless of SES.

Based on the qualitative analysis, quantitative analysis was further carried out. The results suggest that higher SES amplifies the risk of obesity associated with exposure to air pollutants. As shown in Table 4, the results showed that compared with low SES-Q1, there was a synergistic amplification effect of medium and high SES levels and high concentration of NO3− on the risk of overweight/obesity, and the RERI with 95%CI were: 0.723 (0.473–0.973) and 0.562 (0.268–0.856) (all P<0.05). There was a synergistic amplification effect of high concentration of OM and medium level of SES on the risk of overweight/obesity, and the RERI with 95%CI was: 0.672 (0.347–0.997) (P<0.05).

Fig. 3
figure 3

Interaction between air pollutants and SES on overweight/obesity (The adjustment factors were the same as the model 2). The Linear predictor represents the risk of overweight/obesity as predicted by the model.

Full size image
Table 5 RERI (95%CI) of PM2.5 and its components (SO42-, NO3-) and SES on the risk of overweight/obesity.
Full size table
Table 6 RERI (95%CI) of PM2.5’s components (NH4+, OM, BC ) and SES on the risk of overweight/obesity (continued).
Full size table

Discussion

In this study, a longitudinal study of the Jinchang cohort population revealed the association between exposure to PM2.5 and its components, SES levels, and the risk of overweight/obesity. Exposure to PM2.5 and its components increased the risk of overweight/obesity. At different levels of SES, the risk of being overweight/obese increased with increasing pollutant concentrations, and this effect was most significant at the highest levels of SES. There is a synergistically amplified interaction between socioeconomic status and exposure to PM2.5 and its components on the risk of obesity.

Exposure to PM2.5 and its components could increase the risk of overweight/obesity

Exposure to PM2.5is associated with an increased risk of obesity32and induces obesity behaviors such as increased consumption of trans fatty acids and fast food33. The potential mechanisms by which air pollutants increase the risk of obesity may include inflammation, oxidative stress, metabolic imbalance, intestinal flora disorder, and epigenetic modification20,34. In a cell experiment, lipophilic organic chemicals (OC) in PM2.5particles induced inflammation-related genes and increased secretion of the chemokines interleukin (IL)−8/CXLC8 and matrix metalloproteinase 1(MMP1). The expression of aryl hydrocarbon receptor regulatory genes cytochrome P4501A1 (CYP1A1), CYP1B1, and plasminogen activator inhibitor-2 (PAI-2) was significantly up-regulated, while β-adrenergic receptor was down-regulated in primary human adipocytes. These results indicated that OC could interfere with gene expression in adipose tissue35. Air pollutants may cause metabolic disorders by affecting the circadian clock of organisms. In an animal experiment study, it was found that the expression of circadian clock-related genes in two adipose tissues of mice exposed to PM2.5 changed at different time points, suggesting that PM2.5 exposure could destroy circadian rhythms in adipose tissues, which may be an important way of metabolic disorders caused by PM2.5exposure36. Our findings from a population-based experiment showing an inverse association between PM10exposure and reduced telomere length (TL) in overweight or obese persons shed light on a potential mechanism underlying the increase in age-related diseases associated with air pollution exposure37.

Association among exposure to PM2.5 and its components, SES levels, and overweight/obesity

This study found that after adjusting for SES, the risk of developing overweight/obesity increased with increasing pollutant concentrations, and this effect was most pronounced at high SES.

Further analysis found that there was an interaction between socioeconomic status and air pollutants, and higher socioeconomic status amplified the risk of obesity caused by air pollutant exposure.

Some studies have found a positive association between SES and the risk of obesity in developing countries and a negative association between SES and the risk of obesity in developed countries14,15. A study quantifying the socioeconomic inequality of adult obesity in western Iran found that people of higher socioeconomic status, urban residents, and married individuals were more likely to be obese38. A study conducted in Tianjin, China, showed that gender has an impact on the association between socioeconomic status and obesity, with monthly income and education negatively associated with the risk of abdominal overweight/obesity in women, while education was positively associated with the risk of general overweight/obesity in men39.

A Swedish study has demonstrated that individuals with a lower socioeconomic status are subjected to more elevated levels of air pollution40. Concurrently, a research conducted in China has uncovered that the health consequences of air pollution vary among different socioeconomic status groups. Among those of a lower socioeconomic status, self-rated air pollution exerts the most substantial influence on self-rated health. As the socioeconomic status ascends, the impact of self-rated air pollution on self-rated health diminishes24. Collectively, these findings strongly imply that air pollution severely impinges upon the marginalized population with a lower socioeconomic status, exacerbating the existing health disparities and social inequalities25. There are few studies on the effects of socioeconomic status and air pollution on obesity. A study in a rural cohort in Henan, China, found that low socioeconomic status exacerbated the association between air pollution and obesity41, which is contrary to the results of this study and maybe since the study population in this study were all from Jinchang Group Co., LTD. This company is a large non-ferrous metallurgy and chemical industry conglomerate. Most high-socioeconomic-status individuals here are technical managers. They do more mental and less physical work than front-line workers, increasing obesity risk in polluted environments. However, high SES also affects obesity in other ways. For example, those with higher SES have better access to high-calorie diets, like processed and fatty/sugary foods42,43.

Limitations

This study explores the association among atmospheric PM2.5 and its component exposure, SES levels, and population overweight or obesity based on a large cohort population.

This study has some limitations. First, because of the limitations of air pollutants data collection, individual exposure is only based on the air pollutants level measured at the nearest monitoring site of an individual’s residence, without considering the exposure caused by population mobility. Therefore, it is impossible to measure the exposure to particulate matter in the working environment and outdoor activities, which may lead to underestimation of the health effects of PM2.5 and its components. Secondly, the definition of obesity is only the single index of BMI, without considering the type of obesity, such as abdominal obesity, etc., which makes the research results relatively single. In the future, we may further refine measures of pollutant exposure such as incorporating personal air quality monitors or mobile phone tracking data and add other measures of obesity. Finally, although we adjusted for potential confounders, unmeasured residual confounding remains, including greenery status, traffic-related pollutants, and noise exposure.

Conclusion

Summary of findings

Our study demonstrated a significant association between air pollutants (PM2.5 and its components) and overweight/obesity. Higher concentrations of PM2.5 and its components were related to increased risks of overweight/obesity. The risk also varied across different SES levels, with a more pronounced effect at high SES. Specifically, as the concentration of PM2.5 and its components increased, the risk of overweight/obesity escalated, and SES was found to have a modulating effect on this relationship. For example, the hazard ratios (HRs) for different quartiles of PM2.5 and its components in relation to overweight/obesity were consistently significant, and the effect of SES was evident in the differential HRs among SES groups. Synergistic amplification effects were also observed between certain SES levels and specific pollutants such as NO3− and OM on the risk of overweight/obesity, as indicated by the relative excess risk due to interaction (RERI) values.

Implications and future research directions

The findings of this study have important implications for public health. Understanding the complex interplay between air pollution, SES, and obesity can help in devising more targeted interventions. Future studies could focus on exploring the underlying mechanisms through which SES modulates the air pollution-obesity link. Longitudinal studies could be conducted to better understand the temporal sequence of these associations. Additionally, research on the effectiveness of different intervention strategies is warranted. For example, to reduce exposure to pollutants, we could encourage the installation and use of high-efficiency air purifiers in workplaces and homes of high SES individuals with subsidies or tax incentives, promote green transportation options like electric vehicles among them by offering preferential facilities, and implement stricter zoning regulations. For interventions targeting high-SES groups, customized health and wellness programs combining stress management and healthy lifestyle guidance could be designed and offered through exclusive channels, public awareness campaigns could be initiated via social media influencers and high-profile events, and incentives could be provided for their participation in community-based environmental projects. Moreover, further investigations could expand the scope of pollutants and SES indicators to provide a more comprehensive understanding of this relationship.