We systematically searched the PubMed and EMBASE databases using Medical Subject Headings (MeSH for PubMed and Emtree for EMBASE) and relevant keywords with the following strategy before August 14, 2024. Search terms included “Systemic lupus erythematosus,” “Air Pollutants,” and “Environmental Exposure.” The complete search strategy is detailed in Supplementary Material Tables 1 , 2 .

Studies were included based on the following criteria: (1) prospective or retrospective cohort designs or case-control studies; (2) investigations of the association between SLE and air pollution; (3) primary outcomes relating air pollution or environmental exposure to SLE risk, reported as hazard ratios (HR), odds ratios (OR), or relative risks (RR) during follow-up; (4) publications in English.

Exclusion criteria included: (1) duplicate studies; (2) reviews, conference abstracts, comments, or letters; (3) incomplete data or lack of outcomes of interest; (3)Time series study evaluating air pollution and the risk of hospitalization for SLE.

Study selection was performed by two reviewers (YL Xu and HJ Pan), who independently screened literature according to the eligibility and exclusion criteria. If results were consistent, the final analysis proceeded; discrepancies were resolved by consulting full texts and discussing within the group.

Data extraction was conducted independently by the two reviewers (YL Xu and HJ Pan) using Microsoft Excel to create a data extraction table. This table captured information such as the first author, publication year, study type, sample size, follow-up duration, age, SLE diagnosis, and adjusted confounders. Extracted data were cross-checked, and discrepancies were resolved through group discussions.

The Newcastle-Ottawa Scale (NOS) was employed to assess the quality of included studies. The NOS ranges from 0 to 9 points, with higher scores indicating better quality. It evaluates participant selection (4 points), comparability between groups (2 points), and exposure measurement (3 points). NOS scores of ≥7, 4–6, and 0–3 are classified as high, medium, and low quality, respectively.

The adjusted OR and 95% CI from each study were utilized to assess the association between air pollution, emerging chemical pollutants and SLE. We selected random or fixed effects models based on heterogeneity test results. Heterogeneity was considered high if P < 0.1 or I² > 50%. Due to the inherent heterogeneity in meta-analyses—clinical, methodological, and statistical—we employed a random effects model. To explore the sources of heterogeneity, we conducted a priori subgroup analyses based on predefined variables, including study design (cohort vs. case-control), age (adult vs. non-adult), population characteristics (Asian vs. non-Asian populations), and diagnostic criteria. These variables were selected based on clinical plausibility and prior evidence of their potential impact on the risk of SLE. For each subgroup, pooled risk ratios (RRs) with 95% confidence intervals (CIs) were calculated using the random-effects model. Heterogeneity within each subgroup was evaluated by the I² statistic, and differences between subgroups were assessed by comparing the magnitudes of effect sizes and the overlap of confidence intervals. Sensitivity analyses were conducted to evaluate the robustness of results, and a case-by-case elimination method was used to explore sources of heterogeneity. Subgroup analyses were performed based on air pollution type, age, and study type. Publication bias was assessed using funnel plots and Egger’s regression tests.

The initial literature search identified 3,931 pertinent documents, comprising 1,362 articles from PubMed and 2,569 from EMBASE. These records were imported into Note Express reference management software. Upon screening titles and abstracts, irrelevant records were excluded, resulting in 54 articles for full-text evaluation. Ultimately, 8 studies met the eligibility criteria (21–28) and were included in the analysis, as depicted in Figure 1 .

This meta-analysis encompasses 8 studies involving a total of 1,390,348 individuals, published between 2018 and 2024. Among these, four studies were retrospective cohort studies and four were case-control studies. The primary characteristics of the included subjects are outlined in Table 1 . All eight included studies (21–28) reported associations between air pollution exposure and SLE. The diagnostic criteria for SLE varied across studies, with most using established classification systems (ACR or ICD criteria), while one study (26) did not explicitly report the diagnostic criteria used.

The average quality score across the studies was ≥ 6 points, comprising one studies scoring 6 points, three studies scoring 7 points, three studies scoring 8 points, and one study scoring 9 points. This distribution underscores the high quality of research incorporated in this meta-analysis. The NOS scoring scale is detailed in Table 2 .

Visual inspection of the funnel plots revealed a rough symmetry, suggesting no significant publication bias in the association between SLE and air pollution ( Figure 3 ). Consequently, we have a high degree of confidence in the validity of our findings. Furthermore, the Egger regression test yielded a p-value of 0.352 (p > 0.05), reinforcing the absence of publication bias in our meta-analysis.

In this meta-analysis, the moderate heterogeneity observed for PM_2.5 (I² = 56.2%) and PM₁₀ (I² = 66.3%) may stem from multiple factors. In terms of population characteristics, the adult subgroup included studies from different regions and only partial studies controlled for socioeconomic status, while the non-adult subgroup showed 0.0% heterogeneity but was limited by small sample size. Regarding study design, case-control studies (I² = 84.0%) and cohort studies (I² = 88.3%) differed in exposure assessment methods, with the former susceptible to recall bias and the latter’s model estimation potentially ignoring individual activity differences. In terms of exposure windows, the follow-up duration of cohort studies varied widely, and short-term exposure might fail to capture the long latency effect of SLE. Diverse measurement methods for pollutants, such as satellite data, dataset extraction, and spatial interpolation for PM_2.5, also contributed. Additionally, sensitivity analysis using leave-one-out method showed persistent high heterogeneity after excluding individual studies, indicating that heterogeneity was driven by multiple factors including exposure windows, measurement methods, and population differences rather than bias from a single study.

In addition, among the eight included studies, only one study was available for each emerging pollutant exposure, making it impossible to calculate heterogeneity indices for PFASs and BPs.

This meta-analysis included eight cohort studies comprising 1,390,348 participants to investigate the association between air pollution exposure and SLE. The analysis demonstrated significant positive associations between SLE incidence and exposure to PM_2.5 [OR = 1.16, 95%CI (1.02-1.32)] as well as nitrogen dioxide (NO₂) [OR = 1.24, 95%CI (1.11-1.38)]. While these findings provide epidemiological evidence supporting an association between specific air pollutants and SLE, the observational nature of the included studies limits causal inference, and residual confounding cannot be excluded.

A prior meta-analysis (20) focused on the association of PM_2.5 with an increased risk of SLE, highlighting its correlation with SLE Disease Activity Index (SLEDAI) scores. Notably, the time-dependence of PM_2.5’s impact was observed, with no significant association on the third day of exposure but a positive correlation on the sixth day.

In our meta-analysis, encompassing eight articles and 1,390,348 participants, we investigate the association between PM_2.5 and various air pollutants with SLE.Our findings further confirm the significant association between PM_2.5 and SLE. Notably, we observed a pronounced association between NO₂ exposure and SLE. Moreover, we expanded the analysis to include specific chemical environmental pollutants (such as BPs, PFASs), revealing their substantial contribution to SLE.

By incorporating updated research data, our analysis offers a comprehensive assessment of the association between air pollution exposure and SLE, encompassing a wide array of pollutants.

We conducted subgroup analyses based on age and study design. The results showed that the effect size in the non-adult group [OR = 1.58, 95% CI (1.25-1.99)] was higher than that in the adult group [OR = 1.10, 95% CI (0.91-1.33)], suggesting a possible age-related difference. However, it is important to note that the analysis of the non-adult group was based on only two studies with a total of 372 cases, which is insufficient in terms of sample size and number of studies, limiting the reliability of this result. This finding requires further validation in prospective pediatric cohort studies, particularly in conjunction with recent reports of a declining trend in the incidence of childhood SLE (29) for more in-depth exploration. Second, subgroup analysis showed that the association between air pollution and SLE was consistent in both cohort studies [OR = 1.12, 95% CI (0.88-1.42)] and case-control studies [OR = 1.40, 95% CI (0.85-2.29)], although neither reached statistical significance. The effect size observed in case-control studies was larger, a pattern consistent with previously reported methodological differences between observational study designs, suggesting that methodological characteristics should be considered when interpreting assessments of air pollution’s health effects across different study designs.

In our study, although the Egger regression did not suggest significant publication bias (P = 0.352>0.05), the funnel plot showed that small sample studies clustered at the high end of the effect size. This phenomenon implies that there may be a small study effect. Further observation revealed that some of the small-sample studies had relatively low quality scores. This may be related to study design flaws due to small sample sizes, thus increasing the likelihood of the existence of small-sample study effects. For example, 30 patients with childhood-onset systemic lupus erythematosus (cSLE) and 86 healthy controls were selected for the study by Conde, P.G. et al. (21). The sample size was relatively small. This makes the results of the study potentially not representative enough to accurately reflect the relationship between environmental factors and the onset of cSLE in the overall population, and makes it impossible to completely exclude the influence of possible publication bias and small-scale effects of the study. He, G. et al. (23) analyzed the relationship between PFASs and SLE. Their study, included 100 normal and 100 SLE patients, had a limited sample size and was susceptible to the confounding influence of small-scale study effects.

Furthermore, we analyzed the impact of potential confounders that may have been present in the studies. Nearly half of the studies showed that socioeconomic status was a common potential confounder. In general, areas with poorer economic conditions tend to have poorer environmental quality. Residents are more likely to be exposed to high concentrations of PM and have a relatively higher risk of developing SLE. Also, economic conditions may influence residents’ diet and healthcare, which may indirectly affect the onset of SLE. Of the 8 studies included in this review, 4 controlled for socioeconomic status, making the results relatively robust (24–26, 28). However, some studies (21) suggest that occupational exposures or air pollutants of the mother during pregnancy may affect the development of the foetal immune system. In turn, the development of the foetal immune system has been linked to the development of cSLE. This suggests that our genetic susceptibility and occupational exposures may be potential factors influencing the results of the study. However, very few original studies (21) have moderated the two variables of occupational exposure and genetic susceptibility, a factor that contributes to the uncertainty of the study results.

Among the studies we included, air pollution detection methods varied, including environmental monitoring and self-reported exposure. Differences in measurement methods may affect pooled effect estimates to some extent, thereby interfering with accurate judgements of the air pollution-disease relationship. For example, Jung, C. R. et al. (24) used a 1 - km resolution land use regression (LUR) model and a satellite estimation model to estimate air pollutant concentrations. This approach can better reflect the spatial distribution of air pollutants in a large region and effectively capture the differences in pollution in different areas within a city. However, this method may not accurately reflect the actual exposure of individuals due to their different indoor activity times and frequencies; He, G. et al. (23) extracted the residential addresses of patients in their study and obtained air pollutant data with the help of the China Tracking Air Pollution (TAP) dataset. This dataset integrates multiple factors to a certain extent and enables a more comprehensive assessment of air pollutant exposure. These data have been integrated with various factors to some extent, providing a more comprehensive assessment of air pollution exposure. However, the veracity of the data remains susceptible to the influence of various factors, including but not limited to model assumptions and data fusion methodologies. As a consequence, this may introduce biases into the estimates of the pooled effects; Conde, P. G. et al. (21) used a questionnaire to assess exposure. Participants were asked to provide information including the mother’s occupational exposures during pregnancy and the home surroundings. This method of self-reporting exposure is simple and easy to administer, but suffers from recall bias and inaccurate information. Participants may not be able to accurately recall past exposures or may have biased perceptions of certain exposures, leading to errors in exposure data. Therefore, future studies should incorporate multiple methods to improve exposure measures to more accurately assess the relationship between air pollution and SLE and to improve study reliability.

At present, the biological pathogenesis of air pollutants inducing or influencing the occurrence of SLE is not well understood. Several views on its mechanism are mainly related to oxidative stress and immune disorders, and epigenetic alterations (30–32). Air pollution can disrupt helper T cell (Th) homeostasis. It can also activate nuclear factor-κB, which then regulates Th1. Th1 binds to aryl hydrocarbon receptors to regulate Th17 and regulatory T cells. This process triggers the production of pro-inflammatory cytokines (32, 33). This concept is supported by a recent study by Dellaripa et al., which demonstrated that air pollutant exposure can induce specific immune perturbations that are characteristic of early autoimmunity (34). PM, as a representative substance of air pollution, can cause or amplify oxidative stress due to the presence of heavy metals, organic carbon and other complex elements on its surface (35). PM from air pollution, when inhaled, produces oxidants locally in the alveoli, triggering local chronic inflammation. The toxicity of PM depends on its size, shape, and composition, and the transition metals (e.g., Fe, V, Cr, etc.) present in it can produce oxidative stress through a Fenton-type reaction, which can have a negative effect on cells (36, 37). For example, silica particles are toxic to macrophages, inducing cell death and exposing intracellular self-antigens to immune cells (38), which in turn triggers an immune response (39). Inhalation of nanoparticles stimulates alveolar macrophages, triggering an acute systemic inflammatory response, and airway inflammation leads to an increased secretion of pro-inflammatory mediators such as interleukin-8 and granulocyte-macrophage colony-stimulating factor, as well as an influx of neutrophils (40–42). In addition, PM may act as an adjuvant (43), inducing an immune response against otherwise non-immunogenic antigens (44). All of these processes may interfere with normal immune regulation, leading to immune imbalances and increasing the risk of autoimmune diseases. Furthermore, epigenetic modifications play a crucial role in the pathogenesis of SLE (45). Abnormal methylation patterns can lead to aberrant gene activation. For example, hypomethylation of genomic DNA and immune - related genes in CD4+ T cells can cause overexpression of certain genes, thus triggering an autoimmune response (46, 47). Histone modifications, such as acetylation, phosphorylation, ubiquitination, and methylation, affect chromatin structure and gene expression (48). RNA methylation, particularly N6 - methyladenosine (m6A) modification (49), is also implicated in SLE. Downregulation of the demethylase AlkB homolog 5 (ALKBH5) in PBMCs and T cells of SLE patients inhibits apoptosis and promotes T - cell proliferation (50). Upregulation of methyltransferase 3 (METTL3) in the kidneys of SLE patients promotes IRF4 - mediated plasma cell infiltration, resulting in kidney damage (51). In summary, these epigenetic modifications interact with each other and affect the immune system, ultimately driving the occurrence and progression of SLE.

This study conducted a systematic review and meta-analysis to comprehensively assess the epidemiological association between air pollution exposure and the incidence of SLE. Through subgroup analysis, we further explored the specific association patterns between different types of air pollutants (such as PM_2.5 and PM₁₀) and SLE. These findings provide new epidemiological evidence for a deeper understanding of the potential role of environmental factors in the pathogenesis of SLE and offer guidance for future mechanistic studies.

Despite the significant contributions of this study, several limitations warrant consideration. Firstly, the restriction to PubMed/Embase searches might raise concerns about incomplete literature coverage. Although our meta-analysis searched only PubMed and Embase, this approach is methodologically justified. Embase complements PubMed by uniquely covering European and Asian journals in air pollution research (52), ensuring geographical diversity. Methodological analyses further show that including additional databases (e.g., Web of Science, Scopus, or Cochrane Library) identifies fewer than 3% more eligible studies in observational meta-analyses (53, 54). Given the high specificity of our topic—air pollution and SLE—this marginal increase is methodologically insignificant. Our study aligns with the AMSTAR 2 tool’s recommendation (55) to prioritize ‘reasonable’ database selection over exhaustive searches.

Second, Although SLE exhibits well-documented sex disparities in prevalence and clinical manifestations, our meta-analysis could not perform a robust sex-stratified analysis due to insufficient reporting of sex-specific associations between air pollution exposure and SLE in the included studies. Despite our initial intention to explore potential effect modification by sex, the original studies either lacked stratified outcome data or did not explicitly examine interactions between air pollution and sex, precluding meaningful subgroup comparisons. We hope that future research will incorporate more detailed data on sex-specific responses to air pollution, thereby enriching the understanding of how sex may modify the association between air pollution and SLE.

Third, the temporal relationship between pollution exposure and SLE onset exhibited substantial heterogeneity across studies, profoundly influencing the interpretability of exposure-response associations. For instance, studies adopting longitudinal cohort designs (24, 28) evaluated cumulative exposure over 10–15 years prior to SLE diagnosis, aligning with the disease’s postulated latency period (11.77 years median follow-up). Such long-term assessments captured PM_2.5/NO₂ concentrations during critical preclinical phases, identifying threshold effects (e.g., NO₂ 28–38 ppb, PM_2.5 18–46 μg/m³) that correlated with increased risk. In contrast, cross-sectional or case-control studies (23, 27) measured biomarkers (PFASs/BPs) at the time of diagnosis without clear temporal linkage to disease initiation, precluding inference of exposure-onset causality. Studies focusing on pediatric SLE (22) further highlighted temporal complexity, with perinatal exposures (maternal occupational vapor/secondhand smoke) and childhood PM_2.5 demonstrating associations with disease activity and renal failure. However, short-term evaluations (25, 26) using 2-4-year exposure windows (2008-2011) might have underestimated chronic risks, as seen in RA-specific analyses where PM_2.5 exposure 3–5 years prior to diagnosis showed stronger associations (aHR=1.74) than concurrent measurements. This methodological variance introduces bias in pooled effect estimates. Shorter-term assessments (e.g., ≤2 years) in cross-sectional designs (26) may reflect disease-related behavioral changes rather than causal exposure, while self-reported data (21) risk recall bias. For SLE—a chronic disease with protracted preclinical phases—standardizing exposure windows to 5–10 years before diagnosis (28) in future meta-analyses will enhance causal inference, particularly when integrating biological plausibility.

It should also be noted that this meta-analysis has limitations in the geographic distribution of included studies: among all 8 included studies, 6 focused on Asian populations. This lack of diversity in study populations partially limits the generalizability of the findings to other ethnicities and regions (such as North America and Europe). It is important to note that the impact of environmental exposures on autoimmune diseases is mediated by multiple factors, including variations in genetic susceptibility, lifestyle characteristics, baseline disease prevalence, and the specific composition and concentration levels of pollutants in the environment. These factors often exhibit significant regional differences. Therefore, the pooled effect estimates derived from this study may not be directly generalizable to Western populations. This limitation reveals a critical gap in the current literature and the urgent need for large-scale, rigorously designed epidemiological studies in non-Asian populations to examine the links between air pollution, emerging chemical exposures, and SLE. Such studies are essential to validate the reliability of the findings from this research and to refine our understanding within a broader population context.

Additionally, while we identified significant associations for BPs and PFASs in individual studies, the lack of comparable data precluded their inclusion in quantitative synthesis. Future systematic reviews with expanded chemical exposure categories are warranted.