Introduction
Mental disorders are a significant public health issue, and the trends among adolescents are particularly concerning (Fusar-Poli et al., Reference Fusar-Poli, Correll, Arango, Berk, Patel and Ioannidis2021). Specifically, the global prevalence of depression among adolescents has shown a substantial increase, rising from 24% in 2001–2010 to 37% in 2011–2020 (Shorey et al., Reference Shorey, Ng and Wong2022). Notably, research by Ghandour et al. revealed that approximately 75% of children and adolescents diagnosed with depression concurrently experience anxiety (Ghandour et al., Reference Ghandour, Sherman, Vladutiu, Ali, Lynch, Bitsko and Blumberg2019). Both depression and anxiety were associated with increased suicide risk, poor academic performance and social withdrawal (de Lijster et al., Reference de Lijster, Dieleman, Emwj, Dierckx, Wierenga, Verhulst and Legerstee2018; Kalin, Reference Kalin2021). Moreover, when left untreated during adolescence, mental health conditions can persist into adulthood, subsequently leading to more severe physical and psychological symptoms and compromised employment (Canals et al., Reference Canals, Voltas, Hernández-Martínez, Cosi and Arija2019; Duagi et al., Reference Duagi, Carter, Farrelly, Lisk, Shearer, Byford and Brown2024). Therefore, given both short-term and long-term adverse effects, identifying and managing risk factors for depression and anxiety has become increasingly crucial.
Epidemiological evidence has suggested that environmental factors, particularly air pollution, play a crucial role in mental health outcomes (Braithwaite et al., Reference Braithwaite, Zhang, Kirkbride James, Osborn David and Hayes Joseph2019). Recent studies have identified PM2.5 exposure as a potential risk factor for psychological disorders, with research indicating a significant association between elevated PM2.5 levels and the risk of depression and anxiety among adolescents (Jiang et al., Reference Jiang, Luo, Zheng, Xiang, Zhu, Feng and Song2023; S. Li et al., Reference Li, Liu, Li, Xiao, Ou, Tao and Wan2024; Smolker Harry et al., Reference Smolker Harry, Reid Colleen, Friedman Naomi and Banich Marie2024). In parallel, lifestyle risk factors such as physical activity, sleep time and dietary habits have been independently linked to mental health outcomes (Ruiz-Ranz and Asín-Izquierdo, Reference Ruiz-Ranz and Asín-Izquierdo2024; Tian et al., Reference Tian, Cole, Bullmore and Zalesky2024). While these lifestyle components are interrelated, most research has focused on their individual effects rather than their collective impact. Evidence has indicated that healthy lifestyle practices may buffer against PM2.5-related health effects, while unfavourable lifestyle factors combined with high PM2.5 exposure may exacerbate health risks synergistically (Li et al., Reference Li, Xie, Wang, Sun, Hu and Tian2023; Yang et al., Reference Yang, Song, Li, Zhang, Yuan, Wang and Yu2021).
Despite mounting evidence linking both PM2.5 exposure and lifestyle risk factors to mental health individually (Hu et al., Reference Hu, Knibbs, Zhou, Ou, Dong and Dong2024; Pu et al., Reference Pu, Zhu, Shi, Wang, Pan, He and Li2024), several key challenges remain. First, while studies have established associations of long-term PM2.5 exposure (Li et al., Reference Li, Liu, Li, Xiao, Ou, Tao and Wan2024; Smolker Harry et al., Reference Smolker Harry, Reid Colleen, Friedman Naomi and Banich Marie2024) and individual lifestyle factors (Ruiz-Ranz and Asín-Izquierdo, Reference Ruiz-Ranz and Asín-Izquierdo2024; Tian et al., Reference Tian, Cole, Bullmore and Zalesky2024) with mental health symptoms across populations, these exposures are typically examined in isolation. This approach overlooks their potential interactions in the development of depression and anxiety. Given the complex interplay between environmental and behavioural determinants in mental health outcomes, studies focusing on single exposures provide limited insight into their joint effects. Second, there is a critical lack of large-scale studies that simultaneously consider PM2.5 exposure and comprehensive lifestyle risk factors in relation to adolescent mental health, particularly in diverse populations such as China. Moreover, existing approaches have largely relied on single cut-off points to define risk, potentially oversimplifying these complex interactions, and have not incorporated multiple threshold analyses to ensure the robustness of the observed joint effects, which is essential for developing targeted prevention strategies.
Therefore, this study aimed to (1) examine the independent associations of PM2.5 exposure and lifestyle risk factors with depression and anxiety among adolescents; and (2) investigate the potential joint effects between PM2.5 exposure and lifestyle risk factors on depression and anxiety. The findings could inform evidence-based policies for environmental protection and public health interventions targeting adolescent mental health.
Method
Study design and population
The Chinese Adolescent Health Survey (CAHS) was conducted using a multi-stage cluster sampling method between April and December 2021. First, China was geographically divided into five distinct regions – eastern, western, southern, northern and central. One representative province (Jiangsu, Guangdong, Yunnan, Gansu and Hubei) was then randomly selected from each of these five regions. Second, from each selected province, two cities were randomly chosen to ensure a representative sample. Subsequently, within each of these cities, one urban area and one rural area were randomly selected. Third, in each area, one rural junior high school, one rural senior high school, one urban junior high school and one urban senior high school were selected. Fourth, 4 to 6 classes were randomly chosen from each grade level, spanning grades 7 through 12. Finally, students enrolled in the chosen classes were invited to participate in our survey voluntarily. A total of 20153 adolescents, aged between 10 and 18 years (M = 15.16, SD = 1.60), were initially recruited for the study. After excluding 48 participants due to fictitious or inconsistent responses, along with 253 participants who failed to complete the questionnaire, the final analysis included 19852 participants. Written informed consent was obtained from all parents or legal guardians before the study. Ethical approval for this research was granted by the Medical Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology (2021-A216).
Assessment for depression
The Patient Health Questionnaire-9 (PHQ-9), a nine-item tool designed to measure the frequency of depressive symptoms, was used to assess depression (Kroenke et al., Reference Kroenke, Spitzer and Williams2001). The PHQ-9 consists of nine items, each rated on a scale from 0 to 3, resulting in a maximum total score of 27 (Kroenke et al., Reference Kroenke, Spitzer and Williams2001). Participants who scored 10 or above on this scale were classified as screening positive for depression (Manea et al., Reference Manea, Gilbody and McMillan2015). The PHQ-9 showed excellent reliability in this study, with a Cronbach’s alpha of 0.901.
Assessment for anxiety
The severity of anxiety symptoms was measured using the Generalized Anxiety Disorder Scale (GAD-7), a seven-item tool specifically developed for assessing anxiety (Spitzer et al., Reference Spitzer, Kroenke, Williams and Löwe2006). The GAD-7 consists of seven items, each scored on a scale from 0 to 3, resulting in a total possible score ranging from 0 to 21. Higher scores on the GAD-7 indicated more severe anxiety symptoms, with a threshold of 10 or above indicating a positive screening for anxiety (Spitzer et al., Reference Spitzer, Kroenke, Williams and Löwe2006). The Cronbach’s alpha for GAD-7 in this study was 0.937.
Assessment for fine particulate matter (PM2.5)
PM2.5 concentrations were obtained from the ChinaHighAirPollutants (CHAP) dataset (https://weijing-rs.github.io/product.html), which provides comprehensive spatiotemporal coverage across China. The dataset was generated using advanced artificial intelligence models that integrate ground-based measurements, satellite remote sensing data and atmospheric reanalysis (Wei et al., Reference Wei, Li, Lyapustin, Sun, Peng, Xue and Cribb2021). The CHAP dataset offers a high spatial resolution of 1 × 1 km and has been validated for accuracy, with a cross-validated coefficient of determination (R 2) of 0.89 and a root-mean-square error (RMSE) of 10.35 μg/m3 (Wei et al., Reference Wei, Wei, Li, Cribb, Huang, Xue and Song2020). This ensures reliable and precise estimation of daily PM2.5 concentrations for the study period. Based on the geocoded longitude and latitude of each participant’s school address, we assigned daily average PM2.5 exposure levels for the year before the survey to each participant. All students within the same school were assigned uniform PM2.5 concentration levels. The average concentrations were then calculated over 1 year for analysis.
Lifestyle risk factors
A healthy lifestyle score was developed based on eight lifestyle risk factors previously documented at enrolment (CDC, 2025; American Academy of Pediatrics, 2001; Ma et al., Reference Ma, Wu, Shen, Wang, Wang and Hou2022; Gradisar et al., Reference Gradisar, Gardner and Dohnt2011; Xu et al., Reference Xu, Zhang, Wang, Xie, Zhang, Xu, Wan and Tao2023). These factors included smoking, alcohol consumption, physical activity, screen time, sleep duration, dietary behaviours, takeaways and fast food, as well as sugared drinks.
Based on the American Youth Risk Behaviour Surveillance System (YRBSS), participants who reported no smoking or alcohol consumption were categorized as the low-risk group (CDC, 2025). For physical activity, participants who reported engaging in physical activity for 1 hour or more per day were classified as the low-risk group (Ma et al., Reference Ma, Wu, Shen, Wang, Wang and Hou2022). For sedentary behaviour, participants with screen time <2 hours/day and sleep duration ≥8 hours/day were considered as the low-risk group (American academy of pediatrics, 2001; Gradisar et al., Reference Gradisar, Gardner and Dohnt2011). For dietary behaviour, participants were classified as the low-risk group if they met any of the following criteria: eating breakfast ≥4 times per week (Ma et al., Reference Ma, Wu, Shen, Wang, Wang and Hou2022), takeaways and fast food intake ≤2 times per week (Xu et al., Reference Xu, Zhang, Wang, Xie, Zhang, Xu, Wan and Tao2023), or sugared drinks intake <3 times per week (Xu et al., Reference Xu, Zhang, Wang, Xie, Zhang, Xu, Wan and Tao2023). Each low-risk behaviour factor was assigned a score of 0, while high-risk behaviour was assigned a score of 1. The number of risk lifestyle factors was used as a simple score, ranging from 0 to 8, with higher scores indicating a higher level of unfavourable lifestyle.
Covariates
The covariates for the current study included sex (female or male), region of school (rural or urban), grade (junior high school or senior high school), parents’ education level (primary school or less, junior high school, senior high school, or college or more), single-child family (yes or no), family structure (core family, single parent, combined family, grandparents, or others), family income (CNY per capita per month: 6000∼, 4000–5999, 2000–3999, or ∼1999), average temperature, relative humidity and average normalized difference vegetation index (NDVI), The study period’s average temperature and relative humidity data were sourced from the Copernicus Climate Change Service records (https://cds.climate.copernicus.eu/#!/home). NDVI was extracted from the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA’s Terra satellite (https://ladsweb.modaps.eosdis.nasa), which provides vegetation index products with a spatial resolution of 250 m and a temporal resolution of 16 days. Then, the average NDVI value within a 1500 m circular buffer around each school was extracted to represent the level of greenness around the corresponding school.
Statistical analysis
The characteristics of the participants were described using frequencies (%) for categorical variables and median (IQR) for continuous variables. The chi-square test or Wilcoxon rank sum test was utilized to evaluate the distribution differences between participants who reported depression or anxiety and those who did not, according to various characteristics. The dose-response relationships of PM2.5 with depression and anxiety were described using a restricted cubic spline (RCS) with three knots placed at the 10th, 50th and 90th percentiles of the PM2.5 exposure distribution.
Participants were categorized based on PM2.5 exposure levels (Low/High) using multiple percentile thresholds (<P70/≥P70, <P75/≥P75, <P80/≥P80), with the 75th percentile as primary threshold (widely used in environmental epidemiology). Lifestyle was classified as Favourable/Unfavourable using multiple cut-points (<2/≥2, <3/≥3, <4/≥4), with the baseline threshold of 2 based on Jiang et al. (Reference Jiang, Peng, Liu, Lu, Ni and Guo2024) and extended thresholds to accommodate our more comprehensive assessment of lifestyle factors. To evaluate the joint effects of combined exposure to PM2.5 and lifestyle factors, we then generated several categorical dummy variables with four levels based on the binary variables of lifestyle and PM2.5 exposure, including low PM2.5 with favourable lifestyle, high PM2.5 with favourable lifestyle, low PM2.5 with unfavourable lifestyle and high PM2.5 with unfavourable lifestyle. We performed logistic regression analyses to examine the independent associations of PM2.5 exposure and lifestyle risk factors with the prevalence of anxiety and depression. Additionally, participants who exhibited both depression and anxiety were categorized into the comorbid depression and anxiety group, while those who did not meet this criterion were classified as the non-comorbid group. We first examined the associations between combined exposure (PM2.5 and lifestyle) and depression, anxiety and their comorbidity. To further explore potential interactions between PM2.5 and lifestyle, both multiplicative and additive interaction analyses were performed. The additive interaction was assessed using three indicators: relative excess risk due to interaction (RERI, which measures excess risk beyond the sum of individual effects), attributable proportion due to interaction (AP, which quantifies the proportion of disease attributable to interaction), and synergy index (SI, which indicates the strength of interaction relative to individual effects). A RERI > 0, AP > 0, or SI > 1 indicates a positive additive interaction, suggesting that the joint effects of exposures exceed the sum of their individual effects (Andersson et al., Reference Andersson, Alfredsson, Källberg, Zdravkovic and Ahlbom2005; Knol et al., Reference Knol, VanderWeele, Groenwold, Klungel, Rovers and Grobbee2011). Two primary models were used in the analysis: Model 1 included no covariates, while Model 2 accounted for sex, school region, grade, parental education level, single-child family, family structure, family income, average temperature, relative humidity and NDVI. These covariates were selected based on their significance in baseline characteristic comparisons.
To assess the robustness of our main findings, we conducted various subgroup analyses and sensitivity analyses. For the subgroup analysis, we examined the association of co-exposure to PM2.5 and lifestyle and with depression, anxiety and comorbid depression and anxiety, stratified by sex. For sensitivity analyses, we conducted additional analyses using China’s national ambient air quality standard (35 μg/m3) as a cut-off value for PM2.5 exposure. Using this standard-based cut-off, we performed both independent analyses examining associations between PM2.5 exposure and mental health outcomes, as well as joint effects analyses with lifestyle factors at multiple thresholds. To further address potential confounders, inverse probability of treatment weighting (IPTW) was applied to create a pseudo-population that enhances exchangeability between groups. The analytical models were constructed by incorporating adjustments for covariates.
All analyses were conducted using R software (version 4.3.2), with all statistical tests being two-tailed and significance defined as P < 0.05.
Result
Characteristics of study population
The characteristics of participants in the CAHS are presented in Table 1. Among 19852 participants, the mean age was 15.16 (SD = 1.60), and males accounted for 49.80%. Approximately, 54.42% of the participants were from junior high school, and 45.58% were from senior high school. Overall, 3845 and 3230 participants reported depression and anxiety, respectively, with prevalence rates of 19.37% and 16.27%. The median (IQR) values for PM2.5 (μg/m3), average temperature (°C), relative humidity (%) and NDVI were 25.39 (14.94), 17.80 (6.52), 75.34 (4.37) and 0.34 (0.10), respectively. Additional characteristics can be found in Table 1 and Table S1.
Descriptive characteristics of the study participants and stratified by depression and anxiety
Abbreviation: IQR, interquartile range; NDVI, normalized difference vegetation index.
Associations of PM2.5 exposure and lifestyle risk factors with depression and anxiety
We examined the response-dose relationship of PM2.5 with depression and anxiety using RCS (Figure S1). The results indicated no significant nonlinear dose-response relationship between PM2.5 and depression and anxiety (P for nonlinearity > 0.05) (Figure S1), with trend analysis demonstrating a linear trend (P for trend < 0.05) (Table S2). Logistic regression analyses indicated that PM2.5 concentrations above the 70th, 75th and 80th percentiles were independently associated with an increased risk of both depression and anxiety after adjusting for covariates (Table 2). For depression, the odds ratios (ORs) ranged from 1.08 (95% CI: 0.99–1.17) to 1.28 (95% CI: 1.18–1.40). For anxiety, the ORs ranged from 1.11 (95% CI: 1.02–1.21) to 1.28 (95% CI: 1.17–1.40). In addition, the number of lifestyle risk factors exceeding 2, 3 and 4 was also independently associated with an increased risk of depression and anxiety. For depression, the ORs ranged from 3.06 (95% CI: 2.84–3.30) to 3.42 (95% CI: 3.12–3.75), while for anxiety, the ORs ranged from 2.87 (95% CI: 2.65–3.11) to 3.07 (95% CI: 2.70–3.50) (Table 2).
Associations of PM2.5 exposure and lifestyle risk factors with depression and anxiety
Abbreviation: PM2.5, particulate matter with aerodynamic diameter ≤2.5 μm.
Model 1: Crude model without any adjustment.
Model 2: Adjusted for sex, school region, grade, parental education level, single-child family, family structure, family income, average temperature, relative humidity and NDVI.
Unfavourable lifestyle combined with low PM2.5 was used as reference.
* P < 0.05; **P < 0.01; ***P < 0.001.
Note: PM2.5 exposure categories were defined as Low 1/High 1 (<P70/≥P70), Low 2/High 2 (<P75/≥P75) and Low 3/High 3 (<P80/≥P80), where P70, P75 and P80 represent the 70th, 75th and 80th percentiles. Lifestyle categories were defined as Favourable 1/Unfavourable 1 (score <2/≥2), Favourable 2/Unfavourable 2 (<3/≥3) and Favourable 3/Unfavourable 3 (<4/≥4).
Figure 1 displays the distribution of depression, anxiety and their comorbidity across different combinations of PM2.5 and lifestyle thresholds. Participants exposed to high PM2.5 levels combined with unfavourable lifestyle showed increased risks of mental health outcomes (Fig. 2). The strongest associations were found at the 75th and 80th percentiles of PM2.5 with a lifestyle risk score of 4, yielding adjusted ORs (95% CI) of 4.49 (3.79, 5.33) and 4.43 (3.72, 5.28) for depression, 4.01 (3.36, 4.78) and 3.92 (3.27, 4.69) for anxiety, and 4.24 (3.52, 5.10) and 4.23 (3.50, 5.11) for their comorbidity, respectively. No significant multiplicative interactions were observed between PM2.5 exposure and lifestyle on mental health outcomes (all P interaction > 0.05) (Table S3). However, significant additive interactions emerged, particularly at the 80th percentile PM2.5 cut-off with lifestyle scores of 3 and 4 (Table S4). For all mental health outcomes, significant synergistic effects were observed at both lifestyle score thresholds. At score 3, the RERI, AP and SI values were 0.53 [0.09, 1.00], 0.14 [0.02, 0.24] and 1.25 [1.04, 1.49] for depression; 0.47 [0.04, 0.93], 0.14 [0.01, 0.25] and 1.24 [1.02, 1.52] for anxiety; and 0.70 [0.20, 1.24], 0.18 [0.05, 0.29] and 1.33 [1.09, 1.64] for comorbidity. At score 4, these values were 0.85 [0.09, 1.72], 0.19 [0.02, 0.33] and 1.33 [1.04, 1.71] for depression; 0.90 [0.21, 1.69], 0.23 [0.05, 0.36] and 1.45 [1.10, 1.90] for anxiety; and 0.86 [0.09, 1.76], 0.20 [0.02, 0.35] and 1.36 [1.04, 1.80] for comorbidity. Detailed results are presented in Table S4.
Distribution of mental health outcomes using different PM2.5 and lifestyle threshold combinations in adolescents.
Joint effects of PM2.5 concentration and lifestyle on the risk of depression and anxiety and their comorbidity in the total population.
Subgroup analyses
Figure 3 illustrates the sex-stratified associations between combined high levels of PM2.5 exposure and unfavourable lifestyles with mental health outcomes. In males, increasing joint cut-off values of PM2.5 exposure were associated with elevated risks, with adjusted ORs ranging from 1.05 (95% CI: 0.73–1.49) to 4.20 (95% CI: 3.31–5.32) for depression, 1.02 (95% CI: 0.69–1.47) to 4.18 (95% CI: 3.26–5.35) for anxiety and 1.01 (95% CI: 0.64–1.55) to 4.05 (95% CI: 3.06–5.32) for their comorbidity. Similarly, in females, the adjusted ORs ranged from 0.92 (95% CI: 0.63–1.33) to 5.00 (95% CI: 3.89–6.44) for depression, 0.98 (95% CI: 0.65–1.44) to 3.98 (95% CI: 3.09–5.13) for anxiety and 0.65 (95% CI: 0.37–1.07) to 4.56 (95% CI: 3.51–5.07) for their comorbidity. Detailed results are presented in Tables S5 and S6.
Joint effects of PM2.5 concentration and lifestyle on the risk of depression and anxiety and their comorbidity stratified by sex.
Sensitivity analyses
Sensitivity analyses using China’s national ambient air quality standard (35 μg/m3) as the PM2.5 cut-off value yielded results consistent with our main threshold-based analyses (Table S7–S8). IPTW was employed to account for confounding factors across the nine exposure groups, which were defined based on varying thresholds of PM2.5 concentrations combined with different cut-offs for the number of lifestyle factors. Balance diagnostics showed improved covariate distribution after weighting, though some imbalance remained (Table S9–17). The balance test results demonstrated that the matched model exhibited greater robustness (Figure S2). Further logistic regression analysis of the balanced model revealed that any high-level PM2.5 exposure combined with an unfavourable lifestyle was consistently strongly associated with depression, anxiety and their comorbidity (P all <0.05) (Table S18).
Discussion
In this nationwide survey of Chinese adolescents, both high PM2.5 exposure and unfavourable lifestyle were independently associated with increased risks of depression and anxiety. The combination of high PM2.5 exposure and unfavourable lifestyle showed synergistic effects on these mental health outcomes, with associations remaining robust across different cut-off values.
Our findings align with the majority of existing research, indicating that high-level PM2.5 exposure is associated with an increased risk of depression and anxiety among Chinese adolescents (Jiang et al., Reference Jiang, Luo, Zheng, Xiang, Zhu, Feng and Song2023; Li et al., Reference Li, Liu, Li, Xiao, Ou, Tao and Wan2024). However, findings from a longitudinal twin study indicated that exposure to PM2.5 was not associated with depression and anxiety at age 12, but was associated with depression at age 18 (Roberts et al., Reference Roberts, Arseneault, Barratt, Beevers, Danese, Odgers and Fisher2019). These inconsistent findings across studies may be attributed to methodological heterogeneity in population characteristics, exposure assessment and analytical approaches. Notably, median PM2.5 concentrations in Chinese studies exceeded 20 μg/m3 (Jiang et al., Reference Jiang, Luo, Zheng, Xiang, Zhu, Feng and Song2023; Li et al., Reference Li, Liu, Li, Xiao, Ou, Tao and Wan2024), whereas levels in London remained below this threshold (Roberts et al., Reference Roberts, Arseneault, Barratt, Beevers, Danese, Odgers and Fisher2019). The linear trend observed in our study suggests that there is no apparent threshold below which PM2.5 exposure could be considered ‘safe’ in terms of mental health effects. This finding has important public health implications, as it indicates that any reduction in PM2.5 levels, regardless of baseline concentration, may yield proportional mental health benefits.
Modifiable risk factors associated with mental health have emerged as a significant research domain. Studies have increasingly focused on lifestyle determinants, which have become integral components of comprehensive mental health-promoting practices (Peuters et al., Reference Peuters, Maenhout, Cardon, De Paepe, DeSmet, Lauwerier and Crombez2024). Our findings demonstrate robust association between unfavourable lifestyle with the prevalence of depression and anxiety. Two possible mechanistic pathways may elucidate these associations. First, lifestyle factors, particularly physical activity, sleep duration and dietary patterns, modulate mental health through their direct effects on systemic physiology and neurobiological substrates (TiTian et al., Reference Tian, Cole, Bullmore and Zalesky2024). Second, alterations in inflammatory biomarkers induced by unfavourable lifestyle behaviours may serve as potential mechanisms leading to depression and anxiety (Dias et al., Reference Dias, Wirfält, Drake, Gullberg, Hedblad, Persson and Björkbacka2015; Firth et al., Reference Firth, Gangwisch, Borisini, Wootton and Mayer2020). Given that health behaviours are inherently interconnected, examining isolated lifestyle factors may inadequately capture the complexity of behavioural patterns (Gardner et al., Reference Gardner, Champion, Chapman, Newton, Slade, Smout and Sunderland2023). Our study employed a more comprehensive approach to assess lifestyle risk factors and utilized multiple cut-off points to evaluate the associations of lifestyle risk factors with depression and anxiety. These findings provide supportive evidence for improving lifestyle risk patterns to enhance mental health to some extent.
Our analysis revealed a synergistic interaction between PM2.5 exposure and unfavourable lifestyle on depression, anxiety and their comorbidity. This finding has received some support from previous studies (Li et al., Reference Li, Liu, Li, Xiao, Ou, Tao and Wan2024, Reference Li, Xie, Wang, Sun, Hu and Tian2023; Xue et al., Reference Xue, Huang, Liu, Zhang, Feng, Xu and Xu2021), which demonstrated that mental health impacts of PM2.5 exposure could be modulated by lifestyle risk factors, and that adherence to favourable lifestyle practices may attenuate these adverse effects regardless of exposure levels. A similar study found that only smoking demonstrated a significant moderating effect on the potential impact of PM2.5 exposure on mental health (Yang et al., Reference Yang, Song, Li, Zhang, Yuan, Wang and Yu2021). However, this study found no significant moderating effects for other lifestyle behaviours such as alcohol consumption and physical activity (Yang et al., Reference Yang, Song, Li, Zhang, Yuan, Wang and Yu2021). One possible reason for the inconsistency with our findings could be the limited scope of lifestyle assessments in previous studies. In contrast, our study incorporated a broader range of lifestyle indicators, providing a more comprehensive evaluation. The observed additive interaction between PM2.5 exposure and unfavourable lifestyle may be explained through biological pathways directly relevant to depression and anxiety pathophysiology. First, PM2.5-induced systemic inflammation may potentiate lifestyle-related oxidative stress, creating a vicious cycle, which is implicated in mood regulation (Aseervatham et al., Reference Aseervatham, Sivasudha, Jeyadevi and Arul Ananth2013; Liu et al., Reference Liu, Huang, Song, Zhang, Liu and Yu2023). Second, both exposures may jointly alter gut microbiota composition, which has been linked to mental health via the gut-brain axis (Firth et al., Reference Firth, Gangwisch, Borisini, Wootton and Mayer2020; Singh et al., Reference Singh, Sharma, Pal, Kumawat, Shubham, Sarma, Tiwari, Kumar and Nagpal2022). Third, both PM2.5 and lifestyle factors may synergistically increase oxidative stress through mitochondrial dysfunction and reduced antioxidant capacity, potentially accelerating neuronal damage in mood-regulating brain regions (An et al., Reference An, Liu, Shen, Qi, Hu, Song, Li, Du, Bai and Wu2024; Broman et al., Reference Broman, Davis, May and Park2019; Resende et al., Reference Resende, Fernandes, Pereira, De Pascale, Marques, Oliveira, Morais, Santos, Madeira, Pereira and Moreira2020). The additive interaction effects between lifestyle risk factors and PM2.5 exposure demonstrated robust consistency across different cut-off values, supporting the stability of our findings. Furthermore, sensitivity analyses using IPTW to balance potential confounders yielded consistent associations between exposures (PM2.5 and lifestyle risk factors) and mental health outcomes (depression, anxiety and their comorbidity). Given that unfavourable lifestyle risk factors tend to cluster together and their adverse mental health effects may be cumulative (Gardner et al., Reference Gardner, Champion, Chapman, Newton, Slade, Smout and Sunderland2023), a comprehensive lifestyle assessment likely provides more objective insights than evaluating individual factors in isolation. Consequently, the adoption of multiple healthy lifestyle practices should be promoted as a protective strategy against the adverse mental health effects of PM2.5 exposure.
Subgroup analyses further confirmed the joint effects between elevated PM2.5 levels and unfavourable lifestyle factors on the risk of depression, anxiety and their comorbidity. The observed associations demonstrated consistent trends across both males and females. Although previous studies have reported sex-specific differences in PM2.5-related mental health outcomes (Braithwaite et al., Reference Braithwaite, Zhang, Kirkbride James, Osborn David and Hayes Joseph2019), our findings indicate that unfavourable lifestyle factors amplify the adverse effects of PM2.5 exposure similarly in both males and females. These results suggest that intervention strategies targeting both lifestyle modification and PM2.5 exposure reduction may be equally effective across sex groups.
Strengths and limitations
To our knowledge, this is the first study to evaluate the joint effects of PM2.5 exposure and lifestyle on the risk of depression, anxiety and their comorbidity among adolescents. The primary strengths of our study include a large adolescent sample size and rigorous variable definitions. The implementation of multiple classification approaches for both PM2.5 exposure and lifestyle enhanced the methodological rigour. Moreover, comprehensive subgroup and sensitivity analyses demonstrated the consistency of our findings across various populations and analytical frameworks, supporting the robustness of our results.
Several limitations warrant consideration. First, the cross-sectional design precludes causal inference between exposures and outcomes. Longitudinal studies tracking both PM2.5 exposure and lifestyle changes over time would provide stronger evidence for temporal relationships. Second, PM2.5 exposure assessment based solely on school addresses may introduce exposure misclassification (Jiang et al., Reference Jiang, Luo, Zheng, Xiang, Zhu, Feng and Song2023; Li et al., Reference Li, Liu, Li, Xiao, Ou, Tao and Wan2024), although this limitation is partially mitigated by students typically residing near their schools, a common practice in the Chinese educational context. Future studies incorporating personal exposure monitoring or time-activity patterns could provide more accurate exposure assessments and better characterize exposure variability throughout the day. Third, while self-reported depression and anxiety measures have shown strong concordance with clinical diagnoses (Manea et al., Reference Manea, Gilbody and McMillan2015), potential reporting bias cannot be excluded. Employing standardized diagnostic interviews in future investigations would minimize potential reporting bias. Fourth, despite extensive covariate adjustment, residual confounding from unmeasured factors (such as school bullying, peer relationships and family dynamics) remains possible. Future studies should incorporate more comprehensive assessment of social determinants of mental health, including school-based social stressors, and family history of psychiatric conditions to better account for these potential confounders. Finally, the generalizability of our findings to other populations warrants further investigation.
Conclusions
In this nationwide study of Chinese adolescents, we observed independent associations between elevated PM2.5 exposure, unfavourable lifestyle, and increased risks of depression and anxiety. Furthermore, these factors exhibited significant synergistic effects, with their co-occurrence substantially amplifying psychological risks. Notably, adherence to healthy lifestyle practices demonstrated protective effects by attenuating the adverse mental health impacts of PM2.5 exposure. These findings highlight the potential value of integrated intervention strategies that combine air quality improvement with healthy lifestyle promotion to reduce the burden of depression, anxiety and their comorbidity among adolescents.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796025100292.
Availability of data and materials
Data will be made available on request.
Acknowledgements
The authors thank students who took part in the survey, parents who supported the work, teachers who assisted with the field investigation, and all investigators.
Author contributions
Jie Hu and Wei Hu contributed equally to this work and are joint first authors. Yanmei Zhang can also be contacted for correspondence, email zhangyanmei901229@163.com.
Financial support
This work was supported by grants from the Scientific Research Projects from Wuhan Municipal Health Commission (grant number WY22A04) and the National Natural Science Foundation of China (grant numbers 82173541 & 82373599).
Competing interests
None.
Ethical standards
The study procedures were carried out under the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration. The study was approved by the Medical Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology (2021-A216).
Open access
