Depressive symptoms mediate the associations between air pollution exposure and anemia risk: a nationwide longitudinal study in China

Abstract

Background

Anemia remains a critical global health challenge, affecting 24.3% of the global population. While air pollution and depressive symptoms are recognized as potential risk factors, their combined effects and underlying mechanisms on anemia risk remain underexplored. This study aims to explore the associations between air pollution (PM₁, PM₂.₅ and O₃), depressive symptoms, and anemia risk using data from the China Health and Retirement Longitudinal Study.

Methods

Data from 6,526 participants in the CHARLS database from 2011 to 2015 were analyzed. Air pollution exposure was assessed using geospatial modeling, while depressive symptoms were measured by the CESD-10 scale. Anemia was defined based on hemoglobin concentrations. Logistic regression models, restricted cubic spline analysis and mediating effect analysis were used to examine the relationships between air pollutants, depressive symptoms and anemia risk.

Results

Long-term exposure to PM₁ (OR = 1.12, 95% CI: 1.05–1.19, P < 0.001) and PM₂.₅ (OR = 1.10, 95% CI: 1.03–1.17, P = 0.004) was associated with elevated anemia risk, whereas O₃ exhibited a protective effect (OR = 0.68, 95% CI: 0.64–0.73, P < 0.001). Depressive symptoms mediated 5.6% of the PM₁-anemia association (P = 0.020) and 3.1% of the O₃-anemia pathway (P < 0.001). Stratified analyses revealed stronger PM-related anemia risks in rural residents (vs. urban) and western China (vs. eastern) while O₃’s protective effect was pronounced in populations with lower education or pre-existing digestive diseases.

Conclusion

This study provides novel evidence that depressive symptoms partially mediate the effects of PM₁ and O₃ on anemia, underscoring the intertwined roles of environmental and mental health in hematological outcomes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26587-y.

Introduction

Anemia, the world’s most prevalent blood disorder, poses a significant global health burden. Although the age-standardized global anemia prevalence declined from 28·2% in 1990 to 24·3% in 2021, the absolute number of affected people has risen to ~ 2·0 billion owing to population growth, and the burden is distributed unevenly across SDI quintiles and WHO regions [1]. In western sub-Saharan Africa and South Asia, anemia prevalence among women of reproductive age approaches or exceeds 40–45%, driven by dietary iron deficiency, infectious diseases, and limited healthcare access [1]. In contrast, high-income countries such as the United States report significantly lower prevalence rates, typically below 10%, attributed to better nutrition, healthcare infrastructure, and public health interventions [2]. A study examining the burden of anemia among adults aged ≥ 60 years reported that, across all three survey waves, prevalence was consistently higher in rural than in urban areas and in southern compared with northern China (p < 0.05). In 2002 and 2015, the western region also exhibited higher prevalence than the eastern and central regions (p < 0.05). Anemia prevalence increased with age, with the ≥ 85-year group showing double the prevalence of the 60–64-year group (p < 0.05) [3]. China has the characteristics of a large elderly population and accelerated aging. It is inevitable to face the medical pressure brought by anemia in the elderly [4]. These disparities underscore the need for region-specific strategies in anemia prevention and management.

Air pollution is increasingly recognized as a major global health concern, affecting millions of people across diverse environments. Ambient air pollution was linked to 4.2 million premature deaths worldwide in 2019, highlighting its continued heavy toll on global public health [5]. Additionally, air pollution has been shown to be associated with a wide range of diseases. Beyond the well-documented cardiorespiratory outcomes, a growing body of evidence indicates that air pollution exerts systemic effects across multiple organ systems. For example, a nationwide analysis of 117 million hospital admissions across 252 Chinese cities revealed that short-term increases in PM₂.₅ were associated with admissions for a range of conditions, including anemia, diabetes, gastrointestinal hemorrhage, and renal failure, highlighting the susceptibility of hematologic, digestive, genitourinary, and other physiological systems to particulate exposure [6]. Further supporting the systemic impact of air pollution, a population-based cohort study in southwestern China demonstrated that long-term exposure to complex pollutant mixtures was associated with an elevated risk of incident cardiovascular disorders [7].

In light of these broad health implications, recent environmental health research has increasingly focused on specific pollutants with potent biological activity, particularly fine particulate matter (PM₂.₅, particles with a diameter ≤ 2.5 microns), ultrafine particles (PM₁), and ground-level ozone (O₃) [8]. These pollutants are prioritized due to their widespread prevalence, capacity for deep respiratory penetration and systemic translocation, and their established roles in inducing oxidative stress and inflammation. O₃, a secondary pollutant formed by photochemical reactions, is a strong oxidizing agent known to cause respiratory inflammation and systemic oxidative damage [9]. Despite their recognized hazards, the roles of PM₁, PM₂.₅, and O₃ in hematological outcomes such as anemia remain underexplored and sometimes controversial. For instance, while most studies associate PM₂.₅ with increased systemic inflammation and reduced hemoglobin levels [10], evidence regarding O₃ is more conflicted. Some studies report that O₃ exposure is linked to elevated anemia risk, possibly through oxidative hemolysis or suppressed erythropoiesis [11], whereas others suggest that lower-level O₃ might transiently upregulate antioxidant pathways, potentially conferring protective effects under certain conditions [12]. This inconsistency highlights the need for further investigation into the context-dependent biological impacts of O₃. Given the research gaps and distinct physicochemical properties of PM₁, PM₂.₅, and O₃ (e.g., size, deposition, and mechanisms), this study focuses on these pollutants to clarify their individual and divergent links with anemia risk. This will enhance understanding of how specific pollutants affect hematological health, supporting targeted public health and air quality strategies to reduce air pollution’s burden.

A study revealed that both baseline depressive symptoms and their subsequent changes were significantly linked to an increased risk of anemia among older adults, underscoring the complex interplay between mental health and hematological disorders [13]. Findings from Mendelian randomized studies indicate a link between the presence of depression and heightened inflammatory responses [14]. Elevated levels of inflammation are one of the factors contributing to anemia [15]. Building on these parallel lines of evidence, a recent meta-analysis of 18 epidemiological studies demonstrated that both chronic and acute exposures to ambient air pollution are significantly associated with incident depression [16]. Integrating these observations, we hypothesise that depressive symptoms not only constitute an independent risk factor for anemia, but may also act as a modifiable mediator that partially explains how environmental pollution translates into adverse hematological outcomes. Due to disparities in development levels and environmental awareness, the challenges of aging populations and environmental-related issues in developing countries require heightened attention [17]. A systematic review and meta-analysis exploring the global prevalence of depression among older adults specifically assessed with the Geriatric Depression Scale found a pooled prevalence of 35.1% (95%CI: 30.2–40.4%) and that the prevalence of depression decreased with age within this population [18]. Particular attention should be given to the persistently elevated prevalence of anemia in older adults. The sustained manifestation of these conditions may generate societal burdens through increased mean annual direct healthcare expenditures [1, 19]. Research in this domain therefore carries substantial public health significance for addressing health disparities associated with environmental exposures and population aging dynamics.

These findings suggest potential interaction effects between air pollution exposure and depressive symptoms in influencing anemia risk. However, a paucity of longitudinal research has explored such synergistic mechanisms or evaluated their cumulative impacts on long-term hematological outcomes. This study aims to utilize data from the CHARLS database to conduct cohort study, analyzing the potential associations among air pollution, depressive symptom scores, and anemia.

Materials and methods

Data source

CHARLS is a nationally representative longitudinal survey designed to collect dynamic data on health status, socioeconomic characteristics, and aging-related indicators among Chinese adults aged 45 years and older. Initiated in 2011 by the National School of Development at Peking University, the baseline survey employed multistage probability-proportional-to-size sampling to recruit participants across 150 county-level units within 28 provincial administrative regions. Through biennial follow-up waves, CHARLS constructs a comprehensive interdisciplinary database that provides authoritative micro-level evidence for analyzing health disparities in China’s aging population and evaluating the effectiveness of policy interventions [19]. This study was conducted in accordance with the Declaration of Helsinki. The CHARLS survey project has been approved by the Biomedical Ethics Committee of Peking University (IRB00001052-11015). All participants provided written informed consent to participate.

Air pollution exposure assessment

From 2011 to 2015, a geocoding system was employed to map the residential addresses of study participants, facilitating the evaluation of their long-term exposure to air pollution. Data on air pollutants were sourced from the China High Air Pollutants (CHAP) dataset, which provided measurements of PM₁, PM_2.5, and O₃ beginning in 2011. Ground-level concentrations of these pollutants were estimated using artificial intelligence (AI) algorithms, which combined surface measurements, remote sensing data, atmospheric reanalysis outputs, and model simulations. Validation of the AI-estimated concentrations was conducted through 10-fold cross-validation, yielding R² and root mean square error (RMSE) metrics for daily mean concentrations (Table S1). Annual exposure levels were calculated based on each participant’s residential county. Subsequently, long-term exposure was determined by averaging the annual concentrations of PM₁, PM_2.5, and O₃ prior to the 2015 CHARLS survey.

Outcome assessment

Depressive symptoms were evaluated as a potential mediator in the relationship between air pollution and anemia. In the CHARLS baseline survey, these symptoms were measured using the CES-D-10—a validated 10-item self-report scale widely used among China’s older population. CESD-10 is a validated self-report questionnaire used to quantify depressive symptoms over the past week. Each item is rated on a four-point Likert scale (0–3), yielding a total score ranging from 0 to 30, with higher values indicating greater severity of depressive symptoms. A total score ≥ 10 is conventionally interpreted as signaling possible depression and should prompt confirmatory clinical assessment [20]. Participants rated how frequently they experienced specific feelings over the past week on a total scale of 0–30 (with higher scores indicating more severe depressive symptoms) [21–23]. Using a nationally representative subsample of 742 community-dwelling adults aged ≥ 60 years from the 2011 CHARLS wave, Chen & Mui showed that the 10-item Chinese CES-D possesses good internal consistency (Cronbach’s α = 0.80) and a robust two-factor structure (depressive-affect/somatic & positive-affect) that fits the data well (CFI = 0.99, TLI = 0.98, RMSEA = 0.045) [23]. All factor loadings exceeded the 0.30 salience threshold and were invariant across gender, supporting the scale’s reliability and factorial validity for measuring depressive symptoms in mainland Chinese older adults. CHARLS also collected blood samples during both the 2011–2012 baseline and the 2015 follow-up surveys, with anemia defined as a hemoglobin concentration below 13 g/dL for men and 12 g/dL for women [4, 24].

Covariates

This study incorporated a comprehensive set of variables, including demographic characteristics, socioeconomic indicators, lifestyle habits, and health-related measures, to evaluate potential associations between air pollution exposure and anemia. Key demographic factors consisted of participant age, sex, body mass index (BMI), and residential classification (urban versus rural). Socioeconomic parameters encompassed geographic region (categorized as Eastern, Central, Western, or Northeastern China), educational attainment (grouped as primary education or lower, middle school or above), marital situation (married or single), and employment status (retired or not retired). Lifestyle behaviors such as tobacco use and alcohol intake were documented dichotomously (present/absent). Health status was self-rated and classified into three tiers: excellent/good, fair, or poor/very poor. Comorbidities including metabolic (hypertension, diabetes, dyslipidemia), cardiovascular (heart disease, stroke), respiratory (chronic lung disease, asthma), musculoskeletal (arthritis), neuropsychiatric (cognitive/mental disorders), and organ-specific diseases (liver/kidney/digestive disorders, malignancies) were assessed as potential confounders/effect modifiers.

Statistical analysis

Baseline characteristics were summarized with continuous variables expressed as mean ± standard deviation (SD) and categorical variables as frequencies (percentages). The associations between air pollution exposures (PM₁, PM₂.₅ and O₃) and anemia risk were evaluated using logistic regression models. Initially, unadjusted models (Model 1) were constructed; subsequently, Models 2 and 3 were adjusted for potential confounders including age, sex, region, residence, marital status, education level, BMI, smoking, drinking, and overall health status. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated per interquartile range (IQR) increment in pollutant concentrations.

Restricted cubic spline (RCS) analysis was conducted to assess potential nonlinear relationships between pollutant exposures and anemia risk, allowing for a detailed exploration of risk variation across exposure levels. Subgroup and interaction analyses were performed to examine whether these associations differed across demographic and clinical variables. In addition, mediation analyses were applied to determine the extent to which depressive symptoms (assessed via CESD-10 scores) mediate the relationship between air pollution and anemia, with indirect, direct, and total effects estimated and the proportion mediated calculated.

Sensitivity analyses were conducted to evaluate the robustness of our methodological approach. Specifically, we reanalyzed the associations using continuous exposure variables, repeated the analyses after excluding participants with poor or very poor baseline health, and performed separate analyses excluding individuals with cognitive and mental disorders. All sensitivity analyses employed the same model specifications and covariate adjustments as the primary analysis, with significance set at a two-tailed P-value < 0.05. Data analysis was performed using R software (version 4.2.2).

Result

Study population: inclusion criteria and baseline characteristics

Data were derived from the 2011 and 2015 CHARLS waves. From the baseline 2011 cohort of 17,708 participants, we excluded individuals who did not complete the baseline blood collection (n = 6,178), those lost to follow-up or missing repeat blood tests in 2015 (n = 4,106), and participants with pre-existing anemia identified in 2011 (n = 898), yielding a final analytical cohort of 6,526 individuals. The analytical cohort comprised 6,526 eligible participants (Figure S1). Figure 1 displays the geographical representation of the study population across 28 provincial-level administrative divisions in China, demonstrating the nationally representative sampling framework. Baseline sociodemographic and health-related characteristics of the study population are presented in Table 1, including stratified analyses of key covariates.

Fig. 1.

Geospatial Distribution of Anemia Prevalence Across Chinese Provinces

Table 1.

Basic characteristics of participants

Characteristics	Total (n = 6,526)	Non-anemia (n = 5,545)	Anemia (n = 981)	P
Age, (years, M ± SD)	58 ± 9	58 ± 9	60 ± 10	< 0.001
BMI, (kg/m2, M ± SD)	24.6 ± 32.9	24.8 ± 35.1	23.4 ± 15.5	< 0.001
Region, n (%)				< 0.001
Central	1,089 (16.7%)	909 (16.4%)	180 (18.3%)
Eastern	2,950 (45.2%)	2,519 (45.4%)	431 (43.9%)
Northeast	452 (6.9%)	417 (7.5%)	35 (3.6%)
Western	2,035 (31.2%)	1,700 (30.7%)	335 (34.1%)
Night sleep, (hour, M ± SD)	6.37 ± 1.88	6.39 ± 1.86	6.25 ± 1.94	0.043
CESD-10, (M ± SD)	9 ± 6	8 ± 6	10 ± 6	< 0.001
Sex, n (%)				< 0.001
Female	3,502 (53.7%)	2,905 (52.4%)	597 (60.9%)
Male	3,024 (46.3%)	2,640 (47.6%)	384 (39.1%)
Residence, n (%)				< 0.001
Rural	4,319 (66.2%)	3,614 (65.2%)	705 (71.9%)
Urban	2,207 (33.8%)	1,931 (34.8%)	276 (28.1%)
Marital status, n (%)				0.154
Married	5,832 (89.4%)	4,968 (89.6%)	864 (88.1%)
Single	694 (10.6%)	577 (10.4%)	117 (11.9%)
Health status, n (%)				< 0.001
Fair	3,245 (49.7%)	2,761 (49.8%)	484 (49.3%)
Good/very good	1,458 (22.3%)	1,281 (23.1%)	177 (18.0%)
Poor/very poor	1,823 (27.9%)	1,503 (27.1%)	320 (32.6%)
Smoking, n (%)	2,584 (39.6%)	2,273 (41.0%)	311 (31.7%)	< 0.001
Drinking, n (%)	2,707 (41.5%)	2,371 (42.8%)	336 (34.3%)	< 0.001
Education level, n (%)				< 0.001
Elementary school or below	4,570 (70.0%)	3,807 (68.7%)	763 (77.8%)
Middle school or above	1,956 (30.0%)	1,738 (31.3%)	218 (22.2%)
Physical disability, n (%)	1,118 (17.1%)	927 (16.7%)	191 (19.5%)	0.035
Retire, n (%)	600 (9.2%)	526 (9.5%)	74 (7.5%)	0.052
Chronic diseases, n (%)	4,572 (70.1%)	3,872 (69.8%)	700 (71.4%)	0.336
Hypertension,n (%)	3,061 (46.9%)	2,668 (48.1%)	393 (40.1%)	< 0.001
Diabetes, n (%)	970 (14.9%)	832 (15.0%)	138 (14.1%)	0.447
Dyslipidemia, n (%)	739 (11.3%)	652 (11.8%)	87 (8.9%)	0.008
Cancer, n (%)	54 (0.8%)	40 (0.7%)	14 (1.4%)	0.024
Chronic lung disease, n (%)	626 (9.6%)	532 (9.6%)	94 (9.6%)	0.990
Liver disease, n (%)	230 (3.5%)	202 (3.6%)	28 (2.9%)	0.217
Heart disease, n (%)	835 (12.8%)	735 (13.3%)	100 (10.2%)	0.008
Stroke, n (%)	161 (2.5%)	132 (2.4%)	29 (3.0%)	0.284
Kidney disease, n (%)	376 (5.8%)	300 (5.4%)	76 (7.7%)	0.004
Digestive disease, n (%)	1,473 (22.6%)	1,222 (22.0%)	251 (25.6%)	0.014
Mental disorders, n (%)	75 (1.1%)	65 (1.2%)	10 (1.0%)	0.679
Cognitive disorders, n (%)	83 (1.3%)	63 (1.1%)	20 (2.0%)	0.020
Arthritis, n (%)	2,228 (34.1%)	1,840 (33.2%)	388 (39.6%)	< 0.001
Asthma, n (%)	316 (4.8%)	266 (4.8%)	50 (5.1%)	0.687

Abbreviations: BMI body mass index (calculated as weight in kilograms divided by height in meters squared), CESD-10 ​Center for Epidemiologic Studies Depression Scale-10

Compared to non-anemic individuals, anemic participants were significantly older (60 ± 10 versus 58 ± 9 years, P < 0.001) and had a lower BMI (23.4 ± 15.5 versus 24.8 ± 35.1 kg/m², P < 0.001). Significant differences were also observed in regional distribution (P < 0.001) and night sleep duration (6.25 ± 1.94 versus 6.39 ± 1.86 h, P = 0.043). Anemic participants exhibited higher CESD-10 scores (10 ± 6 versus 8 ± 6, P < 0.001), were more likely to be female (60.9% versus 52.4%, P < 0.001) and from rural areas (71.9% versus 65.2%, P < 0.001). Additionally, differences were noted in health status (P < 0.001), smoking (31.7% versus 41.0%, P < 0.001), drinking (34.3% versus 42.8%, P < 0.001), education level (elementary school or below: 77.8% versus 68.7%, P < 0.001), and physical disability (19.5% versus 16.7%, P = 0.035). Among chronic conditions, the prevalences of hypertension (P < 0.001), dyslipidemia (P = 0.008), cancer (P = 0.024), heart disease (P = 0.008), kidney disease (P = 0.004), digestive disease (P = 0.014), cognitive disorders (P = 0.020), and arthritis (P < 0.001) differed significantly between the groups.

Ambient PM₁, PM₂.₅ and Ozone exposure in anemia risk: strengthened particulate effects and persistent Ozone protection across adjustment models

The study examined the associations between long-term exposure to air pollutants and anemia risk using three sequential models with progressive covariate adjustments (Table 2). ​Model 1 adjusted for no covariates, ​Model 2 adjusted for sex and age, and ​Model 3 further adjusted for region, sex, age, residence, marital status, health status, BMI, smoking, drinking, education level, retire status, night sleep duration, CESD-10 score, physical disability, chronic diseases, and specific comorbidities (hypertension, diabetes, dyslipidemia, cancer, chronic lung disease, liver disease, heart disease, stroke, kidney disease, digestive disease, mental disorders, cognitive disorders, arthritis, and asthma).

Table 2.

Associations between air pollution and anemia

Air Pollutions	Model 1			Model 2			Model 3
	OR	95% CI	P	OR	95% CI	P	OR	95% CI	P
PM 1	1.08	1.02, 1.15	0.010	1.07	1.01, 1.14	0.028	1.12	1.05, 1.19	< 0.001
PM 2.5	1.07	1.01, 1.13	0.033	1.06	0.99, 1.12	0.076	1.10	1.03, 1.17	0.004
O 3	0.68	0.64, 0.72	< 0.001	0.67	0.63, 0.72	< 0.001	0.68	0.64, 0.73	< 0.001

Abbreviations: PM₁ particle with aerodynamic diameter ≤ 1 μm, PM_2.5 particle with aerodynamic diameter ≤ 2.5 μm, O₃ ozone; OR Odds Ratio, CI Confidence Interval.

Model 1: no covariates were adjusted

Model 2: adjusted for sex and age

Model 3: adjusted for region, sex, age, residence, marital status, health status, BMI, smoking, drinking, education level, retire, night sleep, CESD-10, physical disability, retire, chronic diseases, hypertension,diabetes, dyslipidemia, cancer, chronic lung disease, liver disease, heart disease, stroke, kidney disease, digestive disease, mental disorders, cognitive disorders, Arthritis, Asthma

In Model 1, PM₁ (OR = 1.08, 95% CI: 1.02–1.15, P = 0.010) and PM₂.₅ (OR = 1.07, 95% CI: 1.01–1.13, P = 0.033) showed significant positive associations with anemia risk, while O₃ exhibited a strong inverse association (OR = 0.68, 95% CI: 0.64–0.72, P < 0.001). After adjusting for sex and age in Model 2, the association for PM₁ remained significant but slightly attenuated (OR = 1.07, 95% CI: 1.01–1.14, P = 0.028), whereas PM₂.₅ became marginally non-significant (OR = 1.06, 95% CI: 0.99–1.12, P = 0.076). O₃ maintained its significant inverse relationship (OR = 0.67, 95% CI: 0.63–0.72, P < 0.001). In the fully adjusted Model 3, PM₁ (OR = 1.12, 95% CI: 1.05–1.19, P < 0.001) and PM₂.₅ (OR = 1.10, 95% CI: 1.03–1.17, P = 0.004) demonstrated stronger and statistically significant associations with anemia risk. Conversely, O₃ continued to show a protective effect (OR = 0.68, 95% CI: 0.64–0.73, P < 0.001), suggesting reduced anemia risk with higher O₃ exposure.

These findings highlight the adverse impact of particulate matter (PM₁ and PM₂.₅) on anemia risk, even after comprehensive adjustments for sociodemographic, lifestyle, health, and comorbidity factors. The consistent inverse association of O₃ with anemia warrants further investigation to elucidate its potential protective mechanisms.

Nonlinear associations between PM₁, PM₂.₅, and O₃ exposure with anemia risk

In this prospective cohort study, we employed RCS analysis with covariates aligned with the fully adjusted logistic regression Model 3 to investigate the nonlinear associations between ambient air pollutants (PM₁, PM₂.₅ and O₃) and anemia risk (Fig. 2). The results demonstrated statistically significant nonlinear relationships for all three pollutants (P overall and P nonlinear < 0.001). Specifically, PM₁ exposure exhibited an inverted U-shaped dose-response curve with a critical threshold at 32.42 µg/m³, where anemia risk initially increased with rising PM₁ levels up to this concentration, followed by a gradual decline at higher exposures. Similarly, PM₂.₅ displayed a comparable inverted U-shaped pattern, characterized by a turning point at 60.23 µg/m³, with risk peaking near this threshold before diminishing. In contrast, the association between ozone exposure and anemia risk showed a multistage pattern: a very gentle upward trend was observed in the dose-response curve until a critical threshold of 76.90 µg/m³ was reached, followed by a transient decline and eventually a significant upward trend (P overall and P nonlinear < 0.001).

Fig. 2.

Nonlinear dose-response relationships between air pollution exposure and anemia risk: restricted cubic spline analysis

Subgroup analysis

Stratified analyses demonstrated significant heterogeneity in the pollutant-disease associations, with differential effects of air pollution on anemia observed across population subgroups characterized by distinct demographic profiles (Fig. 3).

Fig. 3.

Forest plot of subgroup analysis on associations between PM₁, PM₂.₅ and O₃ exposure and anemia risk

Specifically, interactions were identified for residence (urban/rural) and geographic region under PM₁ and PM₂.₅ exposure, as well as educational level, digestive disease status, and geographic region under O₃ exposure. Compared to urban residents (PM₁: OR = 0.92, 95%CI: 0.82–1.04; PM₂.₅: OR = 0.94, 95%CI: 0.84–1.05), rural populations showed significantly elevated anemia risks associated with PM₁ (OR = 1.15, 95%CI: 1.07–1.23; P-interaction = 0.002) and PM₂.₅ exposure (OR = 1.13, 95%CI: 1.05–1.21; P-interaction = 0.007). Geographic disparities were pronounced: participants in western China exhibited markedly higher risks from PM₁ (OR = 2.28, 95%CI: 2.00–2.60) and PM₂.₅ (OR = 2.31, 95%CI: 2.01–2.66) compared to those in eastern China. In contrast, O₃ exposure revealed protective associations among subgroups: individuals with lower educational attainment (OR = 0.65, 95%CI: 0.61–0.70; P-interaction = 0.021), those with digestive diseases (OR = 0.58, 95%CI: 0.51–0.67; P-interaction = 0.011), and residents of central China (OR = 0.60, 95%CI: 0.49–0.74; P-interaction = 0.027).

Mediating effect analysis

To investigate the potential mediating effect of depressive symptoms in the air pollution-anemia relationship, we initially evaluated the epidemiological association between depression severity (measured by CESD-10) and disease outcomes. Table S2 presents the association between CESD-10 and anemia. In the unadjusted model (Model 1), higher CESD-10 scores were significantly associated with increased anemia risk, both as a continuous variable (OR = 1.03, 95% CI: 1.02–1.04, P < 0.001) and in the highest quartile ([12.00–30.00], OR = 1.61, 95% CI: 1.33–1.95, P < 0.001). ​After adjustment for sex and age (Model 2), the association remained significant for both continuous CESD-10 scores (OR = 1.02, 95% CI: 1.01–1.03, P < 0.001) and the highest quartile (OR = 1.43, 95% CI: 1.18–1.73, P < 0.001). However, when fully adjusted for demographic, socioeconomic, and health-related covariates in Model 3, the magnitude of association attenuated - continuous CESD-10 scores showed marginal significance (OR = 1.01, 95% CI: 1.00–1.03, P = 0.044), while the highest quartile of depressive symptoms ([12.00–30.00]) demonstrated non-significant association with anemia risk (OR = 1.22, 95% CI: 0.98–1.52, P = 0.080). This pattern suggests that while depressive symptoms demonstrate an independent association with anemia, their effects may be partially mediated through other health-related pathways captured in the comprehensive covariate adjustment.

Based on these findings, Fig. 4 and Table S3 further explore whether CESD-10 mediates the association between air pollution and anemia. ​For PM₁, CESD-10 demonstrated a statistically significant partial mediation effect, accounting for 5.6% (95% CI: 1.3–18.4) of the total effect (indirect effect coefficient = 0.00064, p = 0.020). In contrast, the association of PM_2.5 with anemia showed no significant mediation through depressive symptoms, with a nonsignificant indirect effect (coefficient=-0.00036, p = 0.080) and negative mediation proportion (-3.6%, 95% CI: -13.7–3.8), indicating that its effects were primarily direct (direct effect coefficient = 0.00912, P = 0.040). Notably, O₃ exhibited a paradoxical inverse mediation pattern: higher O₃ levels were associated with reduced anemia risk through both a small but significant depressive symptom-mediated pathway (indirect effect coefficient=-0.00206, P < 0.001; mediation proportion = 3.1%, 95% CI: 1.6–5.1) and a stronger direct protective effect (coefficient=-0.06408, P < 0.001). These results confirm that CESD-10 statistically mediates the associations of PM₁ and O₃ with anemia, but not PM_2.5, consistent with Table S3’s observation that depressive symptoms partially explain anemia risk in specific exposure contexts.

Fig. 4.

Mediation Effects of CESD-10 in the association between air pollutants (PM₁, PM₂.₅ and O₃) and Anemia

Sensitivity analysis

When air pollution exposures were analyzed as continuous variables (Table S4), fully adjusted models (Model 3) showed significant positive associations between PM₁ (OR = 1.01, 95% CI: 1.01–1.02, P = 0.001) and PM_2.5 (OR = 1.01, 95% CI: 1.00–1.01, P = 0.004) with anemia risk, while O₃ exhibited a protective effect (OR = 0.94, 95% CI: 0.93–0.96, P < 0.001). These associations remained consistent across models adjusting for sociodemographic, behavioral, and clinical covariates.

To address potential confounding by poor baseline health, analyses excluding participants in poor/very poor health (Table S5) strengthened the observed associations for PM₁ (OR = 1.15, 95% CI: 1.05–1.26, P = 0.002) and PM_2.5 (OR = 1.09, 95% CI: 1.00–1.19, P = 0.047) in Model 3. The inverse relationship with O₃ persisted (OR = 0.67, 95% CI: 0.62–0.73, P < 0.001). These findings suggest that the primary associations between particulate matter (PM₁/PM_2.5) and increased anemia risk and O₃ with reduced risk are robust to adjustments for health status and exposure parameterization.

As a further sensitivity check, we repeated the main analysis after dropping participants who had cognitive or mental disorders (Table S6). Compared with the primary fully-adjusted estimates (Model 3), the associations remained almost unchanged: PM₁ (OR = 1.14, 95% CI: 1.06–1.23, P < 0.001), PM₂.₅ (OR = 1.11, 95% CI: 1.03–1.19, P = 0.006) and O₃ (OR = 0.67, 95% CI: 0.62–0.72, P < 0.001). Thus, exclusion of individuals with pre-existing cognitive or mental illness did not attenuate the particulate-matter adverse effects or the ozone protective effect, confirming that the main findings are robust to this potential source of confounding.

Discussion

This study for the first time reported the mediation role of depressive symptoms (CESD-10) in the association between air pollution exposure and anemia risk. In a large-scale population-based analysis, we observed that elevated PM₁ and PM_2.5 levels were associated with increased anemia risk, whereas higher O₃ exposure showed a protective effect. Critically, novel mediation analyses revealed that CESD-10 partially mediated the PM₁-anemia relationship and paradoxically amplified the protective effect of O₃ on anemia through inverse pathways. These findings highlight the complex interplay between air pollutants, mental health, and hematological outcomes, with depressive symptoms acting as a context-dependent mediator in specific pollution-anemia pathways.

This study offers preliminary evidence supporting a potential role of depressive symptoms in linking ambient air pollutant exposure with anemia risk, although the underlying biological pathways remain to be fully elucidated. The direct association between PM₂.₅ exposure and anemia aligns with a nationwide cross-sectional study in India, which reported that each 10 µg/m³ increment in PM₂.₅ was associated to an anemia risk ratio of 1.09 [25]. This association may stem from multiple interconnected mechanisms, primarily driven by PM₂.₅-induced systemic inflammatory responses and oxidative stress that collectively suppress erythropoiesis. Inhaled PM₂.₅ can translocate into systemic circulation, triggering the release of pro-inflammatory cytokines such as IL-6 and TNF-α, which in turn upregulate hepcidin—a key regulator of iron metabolism—leading to functional iron deficiency and impaired hemoglobin synthesis [10, 26]. Supporting this pathway, several studies indicate that systemic inflammation markers such as C-reactive protein (CRP) mediate the relationship between air pollutants and reduced hemoglobin levels [10, 27]. In contrast, ultrafine particulate matter PM₁ appears to operate via a distinct neurotoxic pathway: by penetrating the blood-brain barrier, it activates microglia and triggers the release of IL-6 and TNF-α, thereby reducing the synthesis of monoamine neurotransmitters such as serotonin. This process not only contributes to depressive symptoms but may also indirectly affect erythropoiesis through neuro-immune crosstalk [28]. This neuroinflammatory mechanism may initiate a bidirectional vicious cycle of “depression-anemia comorbidity,” where depression-related appetite suppression leads to insufficient iron intake, while iron deficiency further impairs dopamine synthesis [29, 30]. However, it should be noted that the association between PM_2.5 and incidence risk varies greatly across different regions and within regions, which may be due to differences in the composition of PM_2.5 [31].

Although Peng Zhu’s team reported that PM₂.₅ exposure during mid-to-late pregnancy reduced maternal hemoglobin by 0.94 g/L, increasing the risk of adverse outcomes [32], and Deng’s HAPIN trial found that household PM₂.₅ exposure elevated anemia risk [33]—with iron supplementation mitigating this risk—these findings, obtained in pregnant cohorts, may provide mechanistic insights relevant to other vulnerable groups. Given that elderly individuals are also prone to both air pollution exposure and anemia [10], similar targeted interventions—such as combining iron supplementation (e.g., ferrous sulfate at 60 mg/day) with air purification measures—might represent a promising strategy to mitigate pollution-associated anemia in the elderly.

Notably, O₃ exhibits a biphasic dose-dependent effect. At low concentrations, no significant protective effect is observed; however, in the moderate concentration range, O₃ may exert antioxidant properties through activation of the Nrf2 pathway. This activation can transiently alleviate depression-associated neuroinflammation and consequently reduce anemia risk [12]. Indeed, a study of the Nrf2-mediated antioxidant response to medical ozone confirms that high doses provoke severe oxidative stress and tissue damage, whereas low-to-moderate doses elicit a controlled oxidative eustress that up-regulates antioxidant defenses [12]. It should be noted that some studies have reported ozone as a risk factor for adverse hematological outcomes, likely due to higher exposure levels that induce oxidative stress. Our findings indicate that within a moderate exposure window, the antioxidant effects of ozone may predominate, offering a protective role against anemia. These findings hold significant implications for establishing differentiated air quality standards, suggesting that while the neuroprotective effects of O₃ might be preserved in low-pollution areas, high-pollution regions may require strict implementation of emission reduction policies.

Beyond the direct inflammatory and oxidative pathways, the observed mediation effect of depressive symptoms, particularly for PM₁, suggests the involvement of a more complex neuro-immune-hematological axis. Exposure to ultrafine particulate matter (e.g., PM₁) can induce systemic inflammation and is increasingly linked to neuroinflammatory responses, which are implicated in the pathophysiology of depression [34]. Specifically, such pollutant-induced inflammation can disrupt the kynurenine pathway and monoamine neurotransmitter systems (e.g., serotonin), thereby fostering a neurobiological substrate conducive to depressive symptoms [35, 36]. Crucially, this depression-associated neuroinflammatory state is not confined to the central nervous system but manifests as a systemic, low-grade chronic inflammatory phenotype, characterized by elevated circulating pro-inflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α) [37]. This systemic inflammation is a key driver of “anemia of chronic disease” (ACD, also termed anemia of inflammation). The primary mechanisms involve the induction of hepcidin, the master regulator of iron homeostasis, by inflammatory cytokines (especially IL-6). Elevated hepcidin leads to functional iron deficiency by blocking dietary iron absorption and sequestering iron within macrophages, thereby limiting its availability for erythropoiesis [26]. Concurrently, inflammatory cytokines can directly suppress the proliferation and differentiation of erythroid progenitor cells in the bone marrow and blunt the responsiveness of erythroid precursors to erythropoietin (EPO) [38]. Thus, depressive symptoms act as a significant intermediary, translating ambient PM₁ exposure into a pro-inflammatory state that directly impairs iron metabolism and erythropoiesis, thereby elevating anemia risk.

While this study offers the first population-based evidence that depressive symptoms partially transmit the effects of air pollutants on anemia, its limitations deserve deeper scrutiny. First, the observational design, even fortified by rich covariate adjustment and temporally ordered exposures, still cannot eliminate the possibility of unmeasured or residual confounding; factors such as long-term dietary iron trajectories, genetic polymorphisms affecting heme metabolism, or occult chronic inflammation are absent from CHARLS and could upwardly or downwardly bias observed associations. Second, depression ascertainment rested on a self-reported screening scale rather than on clinical interviews or biological markers (e.g., salivary cortisol, pro-inflammatory cytokines), leaving room for both misclassification and social-desirability bias; likewise, the CESD-10’s somatic items (fatigue, appetite loss) may overlap with early anemia itself, potentially inflating the apparent mediation fraction. Third, although our hybrid satellite–land-use model delivered high-resolution exposure estimates, the lack of personal monitoring or time–activity data prevents us from capturing indoor sources, commuting behavior, and the use of protective devices, all of which could attenuate or amplify true exposure contrasts. Fourth, the 5-year follow-up, typical for longitudinal ageing studies, is too brief to observe multiple anemia cycles or to disentangle acute versus cumulative effects; longer observation could reveal whether the modest 3–6% mediation proportion grows or decays over decades. Future work should therefore embed validated clinical anemia phenotyping, high-frequency personal exposure sensing, and repeated iron-homeostasis biomarkers within pragmatic randomized trials—e.g., testing combined air-purification and targeted iron supplementation—to move from observational inference to mechanistic proof and scalable intervention.

Conclusion

In summary, our study reveals that long-term exposure to PM₁ and PM₂.₅ significantly increases anemia risk, while O₃ appears protective. Depressive symptoms partially mediate these associations, highlighting the interplay between environmental and mental health factors. These findings underscore the urgent need for integrated policies to enhance air quality and public mental health.

Authors’ contributions

Yanqi Kou, Zhe Huang, Shicai Ye, Sitong Yu, Wanyi Yin: Conceptualization, Methodology, Formal analysis, Investigation, Writing – Original Draft. Yanqi Kou, Junyi Yang, Yanxuan Zhong, Ling Qin, Ke Yang, Yujie Huang, Qing Zhang: Data curation, Visualization, Investigation. Rui Ran, Huiqi Huang, Wantong Jiang: Formal analysis, Methodology, Visualization. Yuping Yang: Supervision, Funding acquisition, Writing – Review & Editing. All authors read and approved the final manuscript.

Funding

This study was supported by the Big Data Platform of Affiliated Hospital of Guangdong Medical University, the In-Hospital Funding Clinical Research Project of the Affiliated Hospital of Guangdong Medical University (No. LCYJ2023A002), and the Guangdong Provincial Medical Science and Technology Research Fund Project (No. B2024202).

Data availability

The datasets analyzed during the current study are derived from the China Health and Retirement Longitudinal Study (CHARLS) database, which is publicly available at http://charls.pku.edu.cn.

Declarations

Ethics approval and consent to participate

The China Health and Retirement Longitudinal Study (CHARLS) was approved by the Biomedical Ethics Committee of Peking University (IRB00001052-11015). All participants in the original CHARLS survey provided written informed consent. The current secondary data analysis was exempt from additional ethical approval as it utilized publicly available, de-identified data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.