A comprehensive search on PubMed and EMBASE was used to identify relevant studies, using the keywords (kawasaki) AND/OR (kawasaki disease) AND/OR mucocutaneous lymph node syndrome AND (particulate matter) AND/OR air pollution (Supplementary S1). Due to limited studies on this specific disease outcome, no date restrictions were applied to ensure all relevant studies were included.

This systematic review followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). Citations retrieved from each database were imported into the reference manager Zotero v7.0.11, and duplicate entries were eliminated. Title and abstract screening by two independent reviewers (MK and TJ) determined article eligibility for inclusion. Subsequently, full texts of articles meeting established criteria were further assessed for relevance to this review. Discrepancies in eligibility assessments were resolved through discussion and arbitration by a third reviewer (CM). Studies were deemed eligible for inclusion in the review if they adhered to the following criteria: (1) employed a cohort, case-control, or cross-sectional design; (2) investigated or included the analysis of the prevalence or incidence of Kawasaki disease as an outcome, exposure to particulate matter as a predictor; and (3) incorporated a reference group. Excluded were (1) publications based on cell or animal model data and (2) reviews, commentaries, abstracts, and case reports (Supplementary S2). The protocol for this systematic review was registered with the International Prospective Register of Systematic Reviews (PROSPERO) with registration number CRD42023468937.

Two independent reviewers (MK and TJ) extracted the following information studies that met eligibility criteria: author, study site, design and sample size, objectives, and key findings. Disputes during bias assessments were resolved through discussion and arbitration by a third reviewer (CM). The risk of bias in selected studies was evaluated using the QUADAS-2 tool by NW and MR (Supplementary S3).

Statistical analysis was conducted to compare mean differences in PM₁₀ and PM_2.5 levels between individuals with Kawasaki disease and controls. Given the low heterogeneity among studies as indicated by the I² statistic, fixed-effects (common-effects) meta-analyses were performed only for prenatal and postnatal PM₁₀ levels, and postnatal PM_2.5 levels. The mean difference (MD) was used as the summary measure, with 95% confidence intervals (CIs) calculated for each estimate. Forest plots were generated to visually summarize the meta-analysis results, showing individual study estimates with their weights, as well as the pooled MD and 95% CI. Analyses were conducted using the meta package in R, with the metacont function used to compute pooled effect estimates. Statistical significance was evaluated at p < 0.05. Publication bias was not assessed due to the small number of studies (n = 2). All statistical analyses were conducted using RStudio 2025.05.0 Build 496 (21).

A total of 121 papers met the initial search criteria. After removing duplicates and conducting an initial eligibility screening, 18 papers were selected for full-text review. This review resulted in a final selection of 11 papers that met the full eligibility criteria (Figure 1). These criteria included the use of observational study designs to quantitatively assess exposure to particulate matter (PM₁₀ and PM_2.5), regardless of source; measurable changes in risk, incidence, or prevalence of Kawasaki disease in prenatal and pediatric populations; and inclusion of a reference group (Figure 1).

The methodological quality of the selected studies was assessed using the QUADAS-2 tool. Despite a generally low risk of bias across the domains of subject selection, reference standard, and flow and timing, a high risk of bias was observed in the index test domain. This is partly due to the inherent limitations of environmental exposure studies, which rely on local or regional ambient particulate matter concentrations rather than a specific diagnostic test. Additionally, the selected studies did not specify a threshold but instead used post hoc statistical modeling (Supplementary S3).

A summary of the results is presented in Table 1. Although results varied across studies, positive associations were reported between exposure to particulate matter and increased KD risk. However, only cohort and case-control studies on PM_2.5 showed statistically significant associations. These studies typically involved large sample sizes and longitudinal exposure assessments. In contrast, case-crossover studies assessing short-term exposure showed mixed results, despite similar directional trends.

Due to significant methodological heterogeneity, a meta-analysis of all studies was not appropriate. This heterogeneity stemmed partly from differing outcome measures and regression results. Outcomes were assessed in various ways, with some studies using data from fixed-site monitoring stations and others relying on satellite-based estimations. Predictor variables also varied significantly, with some studies evaluating short-term exposure and others focusing on long-term exposure, leading to notable incomparability when discussing biological mechanisms. Finally, one study notably assessed exposure to suspended particulate matter (SPM) which corresponds to all particles with aerodynamic diameters < 10 µm, with a 50% cut-off at 7 µm, functionally categorizing it as PM₁₀ (22). Nevertheless, comparisons with PM₁₀ should acknowledge this source variability. However, meta-analyses were possible using subset data for pre- and postnatal PM₁₀ exposure (23, 24) and postnatal PM_2.5 exposure (24, 25) as they met criteria and provided data to estimate the mean difference in PM exposure between individuals with Kawasaki and controls.

The pooled analysis of prenatal PM₁₀ exposure levels among children diagnosed with Kawasaki disease compared to controls, based on two eligible studies (n = 43,318), is shown in Figure 2. The pooled mean difference in PM₁₀ concentration was 0.20 µg m⁻³ (95% CI: −0.01 to 0.42), indicating a modest elevation in prenatal exposure among KD cases. Heterogeneity was absent (I² = 0, τ² = 0), suggesting high consistency between studies. However, the result did not reach statistical significance. Similarly, the pooled analysis of postnatal PM₁₀ exposure in Figure 3 shows a mean difference of 0.19 µg m⁻³ (95% CI: −0.04 to 0.46) between children diagnosed with Kawasaki disease and controls (n = 43,318). Heterogeneity was also absent, indicating consistent results. However, the findings were not statistically significant, suggesting that within the 0–10 µg/m³ exposure range, postnatal ambient PM₁₀ levels do not differ meaningfully between cases and controls. Finally, the pooled analysis of two studies evaluating postnatal PM_2.5 exposure between Kawasaki disease cases and controls (n = 24,360) also found no statistically significant difference (Figure 4). The pooled mean difference was 0.95 µg/m³ (95% CI: −0.58 to 2.49). While not statistically significant, the point estimate suggests higher exposure among KD cases.

Inhalation of PM₁₀ particles leads to the penetration of the upper respiratory system and lungs, triggering an inflammatory cascade through oxidative stress and the activation of alveolar macrophages and dendritic cells (26). This activation results in the release of several pro-inflammatory cytokines, including IL-6, IL-1β, and TNF-α which promotes systemic inflammation and oxidative stress (27). Following penetration of the upper respiratory tract, PM₁₀ disrupts tight junctions between endothelial cells leading to increased vascular permeability and in turn may be associated with the vascular inflammation characteristic of KD (28). Despite the biological plausibility, the epidemiological studies included in this review have yielded inconsistent findings on the association between PM₁₀ and KD.

The nationwide longitudinal study in Japan performed by Yorifuji et al. on prenatal and postnatal exposure to suspended particulate matter (SPM) provided supporting evidence for the potential role of PM₁₀ in the development of KD (23). Notably, the ORs for SPM exposure ≥25 µg m⁻³ compared with <20 µg m⁻³ was 1.59 (95% CI 1.06, 2.38). The authors also noted that SPM exposure during mid-to-late gestation appeared to significantly increase the risk of developing KD (23). Despite not reaching statistical significance, the meta-analysis suggests that prenatal exposure to PM₁₀ increase susceptibility of exposed children to KD (23, 24), likely due to its ability to cross the placenta and alter immune system development (Figure 2). Postnatal exposure was also associated with a slight increase in KD risk (Figure 3). Lin et al. in Shanghai, China and Kwon et al. in Korea also showed an increase in KD incidence following exposure to PM₁₀, also did not reach statistical significance (29, 30). Other studies that utilized population data from Taiwan by Jung et al. and Kuo et al. found no significant association between PM₁₀ and KD entirely (24, 31). This suggests that while PM₁₀ may contribute to KD pathogenesis, its impact may be weaker than that of other air pollutants and particulate matter with smaller aerodynamic diameters.

One potential explanation for the weak association between PM₁₀ and KD is the larger size of the particles in comparison to finer particulate matter (PM_2.5) and other air pollutants, which might limit their ability to penetrate deep in the alveolar system and get into systemic circulation (32). As a result, PM₁₀ is generally associated with localized respiratory inflammation rather than the systematic immune response required to induce KD (33). Additionally, variations in study design may have contributed to the weaker association between PM₁₀ and KD. Another crucial limitation across all studies is the lack of genetic stratification, despite KD's well-established link to genetic predisposition. Genetic factors, such as polymorphisms in the CXCL10/IP10 gene, which encodes a chemokine involved in immune function, have been linked to an increased risk of KD, supporting a causal relationship with susceptibility to environmental triggers such as PM₁₀ (34, 35). This suggests that genetic susceptibility may act as a significant modifier, and if not accounted for, studies may underestimate the impact of PM₁₀ on susceptible children.

PM_2.5 is characterized by its ability to penetrate the alveoli and enter the bloodstream, triggering a significant inflammatory response (3, 4). In pooled analyses, postnatal exposure is associated with a slight but not statistically significant increase in KD risk (Figure 4). Despite the debate surrounding the precise etiology of KD, the associated vasculitis is known to involve inflammatory and immune responses, supporting the hypothesis that PM_2.5 may be a significant environmental trigger (36). Buteau et al. found a positive association between ambient PM_2.5 and KD at a Hazard Ratio for interquartile range increments of air pollutant exposure (HR_IQR) of 1.16. Despite not reaching statistical significance, the results also suggested a greater susceptibility of those with preexisting diabetes for KD with respect to the increased inflammatory processes (37). Kim et al. corroborated these findings but with a statistically significant HR of 1.21 for each 5 μg/m³ increase in the annual PM_2.5 concentration and a greater correlation when considering subpopulations (37). They specifically looked at biological sex and found that biological females could be affected more readily by oxidative stress, strengthening the idea that PM_2.5 induces an immune response, which in turn induces KD. Their subgroup analysis showed that for biological females, the HR was 1.39 (95% CI: 1.12–1.73) for each 5 μg/m³ increase in PM_2.5 concentration, while for biological males, the HR was 1.09 (95% CI: 0.90–1.31) (38). This may be partly due to the higher susceptibility of biological females to oxidative stress and greater immune response following exposure to air pollution (39). Additionally, biological females may be more sensitive to air pollutants in early life and more prone to inflammatory responses that could trigger the onset of KD (30).

Kwon et al. found a statistically significant effect of PM_2.5 on KD at lag 1 (one day before the onset of KD), suggesting a possible short-term impact consistent with an immunological trigger (30, 37). Nevertheless, the statistical significance of the daily mean results observed by Kwon et al. and the annual mean results observed by Kim et al. suggest that while short-term PM_2.5 exposure results in an acute inflammatory response, chronic exposure could also lead to prolonged inflammation and immune dysfunction (25). Oh et al. 2021, observed a similar outcome to Kwon et al., with the 2-day average of PM_2.5 concentration showing an OR of 1.01 though no positive associations with the 7-day or 14-day averages (23, 25, 36). Si et al. and Yoneda et al. also found statistically significant increases in KD incidence during high vs. low PM_2.5 seasons, with Yoneda et al. observing a consistent 3%–10% rise in KD incidence for each 1 μg/m³ increase in annual PM_2.5 exposure (17, 19, 37).

Despite variations in statistical significance, all reviewed studies discussed systemic inflammation and immune dysregulation as the primary mechanisms involved in the pathogenesis of KD as a result of both direct inhalation of particulate matter and epigenetic effects. The authors suggest that inhalation of particulate matter is associated with changes in adaptive immune responses and induces cytokine-mediated endothelial cell damage, particularly in genetically predisposed populations (29). A link has also been suggested between perinatal exposure to air pollutants and maternal systemic inflammation, placental inflammatory disorders, and altered fetal immune responses, all of which are implicated in the disease's development. Although statistical significance was not achieved in subgroup analyses of PM exposure alone, findings from multipollutant models highlight the importance of evaluating both synergistic interactions and cumulative exposures when assessing risk (23).

In addition to evaluating multipollutant scenarios, it is also important to consider the social and environmental modifiers of KD incidence. Lin et al. evaluated the impact of high temperatures, described as the 99th percentile (32.4°C), and found a statistically significant increase in KD incidence with a cumulative relative risk of 1.91 (95% CI: 1.13, 3.23) compared to 10°C (37). On the other hand, Yorifuji et al. found that cold weather was also associated with an increased risk of hospitalization, with an odds ratio of 1.51 (95% CI: 1.07, 2.12) for a 20–25 µg/m³ exposure (36). The observed reduction in KD incidence during the COVID-19 pandemic years supports the role behavioral modifiers play in disease incidence. This decline coincided with decreased PM_2.5 levels and the widespread implementation of non-pharmaceutical interventions, including mask mandates and social distancing measures, which ultimately reduced exposure to air pollutants, including particulate matter (37). A similar behavioral influence was observed when adjusting for maternal smoking during pregnancy. Buteau et al. reported a higher incidence rate among children of smoking mothers (19.8 cases per 100,000 person-years) compared to those of non-smoking mothers (12.3), again highlighting the role behavior can play in exposure levels and disease incidence (36). Socioeconomic modifiers also play a significant role in disease incidence. Populations from lower income settings were found to have higher KD incidence likely due to the cumulative burden of inadequate housing and healthcare access, as well as a limited ability to reduce exposure through solutions such as air purifiers (40). Buteau et al. specifically adjusted for neighborhood socioeconomic status and found higher incidence rates in areas of high material deprivation, at 13.4 cases per 100,000 person-years, compared to between 10.1 and 12.9 cases in areas with lower levels of material deprivation (41). These findings emphasize the importance of incorporating social, behavioral and environmental modifiers in future exposure assessment models, thus strengthening risk stratification and intervention targeting.

Although the studies in this review reported an overall positive association between particulate matter exposure and KD risk, several methodological limitations may affect the reliability of these findings. Firstly, the total number of reviewed studies was small, which highlights the need for additional primary research. Secondly, many studies relied on environmental exposure measurements, such as fixed-site or satellite observations, rather than individual-level assessments which may not accurately reflect personal exposure levels (42). Environmental data carries notable causal uncertainty, as generalized exposure measurements may underestimate the true effects (43). These challenges are especially pronounced in air pollution research, where the spatial resolution is often too broad to capture localized variability (44). Another important consideration is the mobility of participants during the study period, which can meaningfully impact analyses of perinatal outcomes, given variations in susceptibility throughout pregnancy and postpartum (24, 25, 31). Finally, while administrative health data help reduce recall bias, they often lack clinical detail and can inflate estimates. This particular challenge, combined with the absence of a gold standard diagnostic test for Kawasaki disease, increases the risk of misdiagnosis or outcome misclassification (45, 46).

Some of the results from the reviewed studies did not reach statistical significance, despite this overall trend of a positive association between increased particulate matter exposure and KD. Thus, as suggested by authors, future research should focus on validating this association across a range of study populations and geographical settings, and examine the biological and pathophysiological mechanisms involved (30). Addressing limitations in exposure assessment through individual-level exposure measurements, which are critical in air pollution research, may also strengthen the results (45, 46). Finally, as noted by Kwon et al., given the limitations of relying solely on clinical features for diagnosis, future studies may benefit from including objective inflammatory markers as intermediate indicators of Kawasaki disease (30).