Prenatal and Postnatal Ambient Air Pollution and Kawasaki Disease

Abstract

Background

Accumulating evidence indicates associations between ambient air pollution and Kawasaki disease (KD), but the results remain inconsistent.

Objectives

This systematic review and meta-analysis aimed to comprehensively summarize the current evidence on the effects of ambient air pollutants on KD.

Methods

The PubMed, Web of Science, Embase, and Scopus databases were searched up to January 18, 2025 for studies investigating the effects of ambient air pollution on KD. A fixed- or random-effects model was used to calculate pooled ORs with 95% CIs for an increase in ambient air pollutant concentration of 10 μg/m³. The risk of bias was assessed using the Risk of Bias In Nonrandomized Studies of Exposures tool, and the quality of evidence was assessed by the Grading of Recommendations, Assessment, Development, and Evaluations framework. The protocol was registered with PROSPERO (CRD42024545321).

Results

Thirteen studies with 124,857 participants were included. Seven studies were at high risk of bias. The meta-analysis revealed an increased risk of KD after short-term postnatal exposure to PM_2.5 (OR: 1.011; 95% CI: 1.003-1.019; I² = 0%; high-quality evidence) and PM₁₀ (OR: 1.004; 95% CI: 1.000-1.008; I² = 38%; high-quality evidence), as well as long-term postnatal exposure to PM_2.5 (OR: 1.415; 95% CI: 1.179-1.697; I² = 41%; high-quality evidence). Prenatal exposure to carbon monoxide, nitric oxide, nitrogen oxides, and sulfur dioxide; short-term postnatal exposure to nitric oxide; and long-term postnatal exposure to carbon monoxide and nitrogen oxides were also associated with KD occurrence.

Conclusions

Both prenatal and postnatal exposure to several ambient pollutants are associated with the risk of KD.

Central Illustration

Kawasaki disease (KD) is a systemic vasculitis that affects mainly children and has been reported across all ethnicities spanning all continents, with the highest rates in East Asia.^1,2 Despite the availability of effective treatments, KD still poses a great risk for coronary artery and myocardial inflammatory injury because of either delayed treatment or aggressiveness of the disease.3, 4, 5 KD is the most common cause of acquired heart disease in children in high-income regions and is associated with a higher risk for cardiovascular events in adulthood, leading to public health concerns.6, 7, 8, 9, 10 The causes of KD are not completely understood, most likely resulting from the interplay between genetic susceptibility and environmental triggers, involving mechanisms related to immune cell activation, exaggerated inflammatory response, and increased oxidative stress.11, 12, 13, 14, 15

Air pollution has become a major contributor to the global disease burden, leading to 8.1 million deaths in 2021, including 709,000 deaths in children under 5 years of age.^16,17 Emerging studies have indicated that exposure to high levels of ambient air pollutants is associated with immune-mediated inflammatory diseases and cardiovascular diseases by triggering innate and adaptive immune responses, inducing inflammation, stimulating oxidative stress, and mediating genetic and epigenetic mechanisms.18, 19, 20, 21, 22, 23 The geographical clustering and seasonal variation of KD incidence and the multiple overlapping mechanisms of KD and air pollutants link KD with environmental factors, including ambient air pollution.^24,25

Owing to a range of biological and behavioral factors, the fetuses, infants, and children are uniquely susceptible to the detrimental effects of ambient air pollution.^26,27 Importantly, comprehensive interventions targeting modifiable environmental risk factors in early life are crucial for cardiovascular conditions across the lifespan.²⁸ In these circumstances, exploring the associations between prenatal and postnatal exposure to ambient air pollution and KD is of potential public health significance.

Although the associations between ambient air pollution and KD have been investigated in recent years, the results have been inconsistent. For example, a study in Korea reported that exposure to PM_2.5 (particulate matter with an aerodynamic diameter ≤2.5 μm) was significantly positively associated with the risk of KD,²⁹ whereas a study in 7 metropolitan regions of North America did not yield this result.³⁰ Zhu et al³¹ indicated an increased risk of KD onset after ozone (O₃) exposure, while Kwon et al³² did not observe this association.

Consequently, we conducted this systematic review and meta-analysis of population-based epidemiological studies to investigate the relationship between exposure to ambient air pollutants in pregnant women and children and the occurrence of KD in children.

Methods

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement (Supplemental Table 1).³³ The review protocol was preregistered in PROSPERO (CRD42024545321).

Data sources and search strategy

Four electronic databases (PubMed, Web of Science, Embase, and Scopus) were searched from inception to 11 May 2024 (initial search) and 18 January 2025 (final search) for relevant studies investigating the risk of KD after exposure to ambient air pollution. No limitations were placed on publication year or language. We used the following search terms: air pollution, air pollutants, particulate matter (PM_2.5, PM₁₀, suspended particulate matter [SPM], fine particulate matter, fine particles, quasi-ultrafine, ultrafine particles), carbon monoxide (CO), nitric oxide (NO), nitrogen dioxide (NO₂), nitrogen oxides (NO_X), sulfur dioxide (SO₂), O₃, black carbon, wildfire, mucocutaneous lymph node syndrome, Kawasaki syndrome, and KD. Manually searching the references of the retained studies was utilized to avoid potentially missed studies in the database searches. Further details of the search strategy are provided in Supplemental Table 2.

Study selection criteria

The PECOS framework (Participants, Exposure, Comparator, Outcome, and Study) was used to formulate the eligibility criteria. Our study participants were pregnant women and children (birth to 19 years). Exposure indicated exposure to ambient (outdoor) air pollutants of anthropogenic or biogenic sources, with metrics established over a certain time period. Exposure during pregnancy was considered prenatal exposure, and exposure between birth and 19 years of age was considered postnatal exposure. Further, we defined short-term postnatal exposure as <1 month and long-term postnatal exposure as ≥1 month.³⁴ Any comparison was included. The investigation outcomes were hospitalization or incidence or prevalence of KD. The diagnosis of KD was performed according to International Classification of Diseases-9, International Classification of Diseases-10, American Heart Association criteria, Japan Kawasaki Disease Research Committee criteria, or clinical diagnosis by physicians. Any type of population-based epidemiological study, including case-crossover, time series, and cohort studies, met the criteria for this study. The detailed inclusion and exclusion criteria are presented in Supplemental Table 3.

After removing duplicates, 2 authors (P.Y. and J.Z.) independently screened studies via 2 phases: 1) all titles and abstracts screening and 2) full-text screening for studies that passed the first screening. Disagreements were resolved through discussion or by consultation with a third author (K.Z.).

Data extraction

Two authors (P.Y. and J.Z.) extracted the information from all eligible studies. The data extracted from the studies included: 1) citation details (first author, publication year, and study period); 2) study design details (geographical location, type of study design, age, male ratio, and sample size); 3) exposure details (exposure assessment method, type of air pollutant, air pollutant level, units of increment, and exposure duration); 4) outcome details (KD), association effect estimates with 95% CIs, outcome types (hospitalization, incidence, or prevalence), and the number of cases; and 5) adjusted covariates (eg, age, sex, and socioeconomic status variables). Disagreements were resolved by discussion until a consensus was reached, and if necessary, a third author (K.Z.) was consulted.

Risk of bias assessment

Two authors (P.Y. and J.Z.) assessed the risk of bias via the Risk of Bias In Nonrandomized Studies of Exposures (ROBINS-E) tool.³⁵ The ROBINS-E consists of 7 domains: 1) bias due to confounding; 2) bias in the selection of participants into the study; 3) bias in exposure classification; 4) bias in departure from intended exposure; 5) bias due to missing data; 6) bias in outcome measurement; and 7) bias in the selection of reported results (Supplemental Table 4). Each domain was rated as low risk of bias, moderate risk of bias, or high risk of bias. The overall risk of bias was recorded as the highest risk of bias in any domain for each study.

Data standardization and synthesis methods

We assumed a linear relationship between each ambient air pollutant and KD and utilized the OR as the effect estimate to evaluate the relationship. Given that the incidence of KD caused by ambient air pollutants is relatively low, the relative risk was considered approximately equal to the OR.³⁶ For studies reporting percentage changes in prevalence, the percentage change was divided by 100 and then increased by 1 to obtain the OR. For studies that reported a regression coefficient (β, ie, natural log relative risk), the regression coefficient was exponentiated to obtain the OR.³⁷

For different lag structures in short-term postnatal exposure, the results for the lag days that were the focus of the study or yielded the largest effect estimate were selected, as no reliable evidence indicated how to choose the optimal lag day. For multiple time points in long-term postnatal exposure, the results for the longest time point were selected. For the single- and multiple-pollutant models, the results of the single-pollutant model were selected. For studies reporting results of the non-COVID-19 and COVID-19 periods, the results of the non-COVID-19 period were selected. For the unadjusted and adjusted models, the results of the final adjusted model were selected. For studies reporting data from industrial and nonindustrial areas, only the risk outcome values of industrial areas were selected for analysis because industrial areas may be associated with higher exposure levels and disease risks.

To facilitate comparisons across studies, we converted the effect estimate of each ambient air pollutant to concentration increments of 10 μg/m³. Conversions from other units to μg/m³ were conducted via the following formula: 1 ppb CO = 1.145 μg/m³; 1 ppm CO = 1,145 μg/m³; 1 ppb NO = 1.23 μg/m³; 1 ppb NO₂ = 1.88 μg/m³; 1 ppb NO_X = 1.9125 μg/m³; 1 ppb SO₂ = 2.62 μg/m³; and 1 ppb O₃ = 1.96 μg/m³. We subsequently recalculated the standardized OR and 95% CI (the conversion of the CI was similar to that of the OR) for each ambient air pollutant in each study via the following formulation: OR (standardized) = OR (original)^{Increment (10)/Increment (original)}.³⁸

Statistical analysis

Meta-analyses were conducted if at least 3 included studies explored the association between the same pollutant and KD. Data were pooled via the fixed-effect model (inverse variance method) or the random-effect model (DerSimonian‒Laird method with Hartung‒Knapp adjustment) according to whether significant heterogeneity existed.³⁹ The heterogeneity was quantitatively evaluated via the Cochran's Q-test and I² statistic, with a P value for the Q test <0.1 or I² >50% considered significant heterogeneity.⁴⁰ For meta-analyses with significant heterogeneity, subgroup analysis (study design, country development, percentage of males, and population size) was conducted to explore the possible sources of heterogeneity. To investigate the robustness of the results, sensitivity analysis was performed on all results that included at least 3 studies by sequentially removing each study. Publication bias was also assessed for analyses of 3 or more included studies by Egger's test and Begg's test.^41,42 The trim and filling method was applied if the P value of Begg's and Egger's tests was <0.05.⁴³

All analyses were conducted using R software (version 4.2.3) with the “meta” package.

Quality assessment

The Newcastle Ottawa Scale (NOS) and the modified NOS were used to assess the quality of the included studies. Specifically, the NOS used for cohort studies measures 3 dimensions: selection (4 items, 1 star per item), comparability (1 item, up to 2 stars), and outcome (3 items, 1 star per item). A study with a NOS score of 1 to 3, 4 to 6, or 7 to 9 was considered poor, intermediate, or high quality, respectively.⁴⁴ The modified NOS used for case-crossover and time-series studies measures 3 domains: quality of pollutant (0-1 point), quality of outcome (0-1 point), and degree of adjustment for confounding (0-3 points). A study with a modified NOS score of 0 to 1, 2 to 3, or 4 to 5 was considered poor, intermediate, or high quality, respectively.⁴⁵

The Grading of Recommendations, Assessment, Development, and Evaluations approach was used to evaluate the quality of evidence for each pair of exposure outcomes.⁴⁶ The quality of evidence was rated as high, moderate, low, or very low by evaluating 8 domains (risk of bias, inconsistency, imprecision, indirectness, publication bias, dose‒response trend, the magnitude of associations, and residual confounding). The details of the Grading of Recommendations, Assessment, Development, and Evaluations assessment are provided in Supplemental Methods 1.

Results

Study selection and study characteristics

A total of 350 studies were identified across the 3 different databases (Figure 1). We removed 181 duplicates, 148 studies by judging the titles and abstracts, and 8 studies through full-text screening (Supplemental Table 5), leaving 13 studies for inclusion.29, 30, 31, 32^,34,47, 48, 49, 50, 51, 52, 53, 54

Figure 1.

PRISMA Flow Diagram of Study Screening and Selection

PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

The characteristics of the included studies are presented in Table 1. The majority originated from East Asia (n = 11), and the most common study design was case-crossover (n = 5). One study evaluated prenatal exposure only, 10 studies assessed postnatal exposure only, and 2 studies investigated both prenatal and postnatal exposure. The age range across the studies was 1 month to 18.8 years. The air pollutant concentrations ranged from 0.190 to 49.090 μg/m³ for PM_2.5, 47.990 to 92.900 μg/m³ for PM₁₀, 20.000 to 20.100 μg/m³ for SPM, 572.500 to 858.750 μg/m³ for CO, 3.038 to 7.724 μg/m³ for NO, 1.140 to 62.300 μg/m³ for NO₂, 46.799 to 48.138 μg/m³ for NO_X, 2.280 to 46.000 μg/m³ for SO₂, and 56.036 to 87.200 μg/m³ for O₃.

Table 1.

Study Characteristics of the Included Studies

Study							Exposure		Exposure		Outcome
First Author (Year)	Location	Design	Time Period	Events	Age	Male (%)	Air Pollutants	Mean Concentration (μg/m3)	Assessment Method	Duration	Outcome	Evidence	Covariates Adjusted
Jung (2017)	Taiwan, China	Case-crossover	2000-2010	695	2.54 ± 1.21 y	56.8	PM10	57.390	Monitoring stations	Short-term after birth	KD admission	ICD-9-CM code 446.1	Temperature, humidity, northward wind, eastward wind
							CO	858.750
							NO2	43.616
							SO2	11.607
							O3	83.790
Kwon (2022)	Korea	Case-crossover	2007-2019	51,486	Under 5 y	58.2	PM2.5	25.040	Monitoring stations	Short-term after birth	KD admission and IVIG treatment	ICD-10 code M30.3	Temperature and humidity
							PM10	47.990
							CO	732.800
							NO2	49.519
							SO2	12.890
							O3	73.304
Lin (2017)	Shanghai, China	Time series	2001-2010	2,344	2.35 y (1 mo to 18.8 y)	69.5	PM10	92.900	Monitoring stations	Short-term after birth	KD admission	Japanese/AHA diagnostic criteriaa,b	Temperature, humidity, and long-term and inherent seasonal trends in KD incidence
							NO2	62.300
							SO2	46.000
Buteau (2020)	Quebec, Canada	Cohort	From birth (2006-2012) until March 31, 2018	539	3 y (0.1-10.4 y)	60.1	PM2.5	0.190	Satellite, land-use regression models, and dispersion models	Prenatal	KD admission	ICD-10 code M30.3	Maternal age, parity, sex, multiple birth, maternal smoking during pregnancy, material deprivation, birth year, and rural/urban residence
							NO2	1.140
							SO2	2.280
Kim (2024)	Korean	Cohort	2002-2019	1,220	Under 5 y	58.4	PM2.5	27.400	Prediction data developed by AiMS-CREATE network	Long-term after birth	KD admission and IVIG treatment	ICD-10 code M30.3	Age, sex, income status of each individual, and community-level socioeconomic indicators, spatial and temporal trends
Yorifuji (2024)	Japan	Cohort	Prenatal: 08/2009-04/2010; Postnatal: 01/2011-12/2012	193	6-30 mo	57.0	Prenatal: SPM	20.100	Monitoring stations	Prenatal and long-term after birth	KD admission	Japanese diagnostic criteriaa	Children's (sex, singleton birth, parity, breastfeeding status, daycare attendance), maternal (age, smoking status, education) characteristics, residential area, and per capita taxable income
							Postnatal: SPM	20.000
Oh (2021)	Seoul, Gyeonggi, and Incheon, South Korea	Case-crossover	2006-2016	771	Under 10 y	59.4	PM2.5	34.130	Ground and phase observations and CMAQ modeling data	Short-term after birth	KD admission	ICD-10 code M30.3	Temperature and humidity
Zeft (2016)	USA and Canada	Case-crossover	1990-2011	3,099	NA	NA	PM2.5	7.000-17.200	Monitoring stations	Short-term after birth	Diagnosed at KD registries	AHA diagnostic criteriab	Temperature and dew point data
Kuo (2022)	Taiwan, China	Cohort	2004-2010	4,192	Under 6 y	62.0	Prenatal:		Monitoring stations	Prenatal and long-term after birth	KD admission and IVIG treatment	ICD-9-CM code 446.1	Age, gender, birth weight, mother age, mode of delivery, preterm delivery, and maternal comorbidity
							CO	595.400
							NO	7.724
							NO2	35.513
							NOx	48.138
							O3	56.036
							Postnatal:
							CO	572.500
							NO	7.724
							NO2	34.611
							NOx	46.799
							O3	56.291
Zhu (2024)	East China	Case-crossover	2013-2020	1,808	2 y (-5 y)	62.0	O3	87.200	Satellite-based model	Short-term after birth	KD admission	ICD-9-CM code 446.1/ICD-10 code M30.3	Temperature and humidity
Si (2023)	Chengdu, China	Time series	2015-2021	3,036	2 y (38 d-13 y)	66.3	PM2.5	49.090	Monitoring stations	Long-term after birth	KD admission	Cutaneous mucocutaneous lymph node syndrome and IVIG nonresponsive KD	Temperature and humidity
Fujii (2020)	Beppu, Japan	Time series	2011-2018	185	2.1 y (1 mo-12 y)	53.5	NO	3.200	Monitoring stations	Short-term after birth	KD admission	AHA diagnostic criteriac	Temperature, humidity, and long-term and inherent seasonal trends in KD incidence
							NO2	14.300
							SO2	11.200
Yoneda (2024)	Japan	Time series	July 2014-December 2019	55,289	Under 5 y	NA	Postnatal:		Monitoring stations	Long-term after birth	KD admission	ICD-10 code M30.3	Long-term trends, seasonality, day of the week, and offset variable: Log of person-days (based on under-5 population from 2020 Census) to standardize incidence rates
							PM2.5	11.400
							NO	3.038
							NO2	14.664
							SO2	3.327

AHA = American Heart Association; CO = carbon monoxide; ICD = International Classification of Diseases; IVIG = intravenous immunoglobulin; KD = Kawasaki disease; NO = nitric oxide; NO₂ = nitrogen dioxide; NO_x = nitrogen oxides; O₃ = ozone; PM = particulate matter; SO₂ = sulfur dioxide; SPM = suspended particulate matter.

^aThe 5th revised edition of diagnostic criteria for KD, issued by the Japan Kawasaki Disease Research Committee - Revision of diagnostic guidelines for kawasaki disease (the 5th revised edition).

^bThe criteria of the 2004 AHA statement - Diagnosis, treatment, and long-term management of Kawasaki disease: a statement for health professionals from the committee on rheumatic fever, endocarditis, and Kawasaki disease, council on cardiovascular disease in the young, American Heart Association.

^cThe criteria of the 2017 AHA statement - Diagnosis, treatment, and long-term management of Kawasaki Disease: a scientific statement for health professionals from the American Heart Association.

Risk of bias

With the ROBINS-E tool, the risk of bias was rated as low in 1 study,⁴⁹ moderate in 5 studies,^{29,32,34,47,51} and high in 7 studies^30,31,48,50,52, 53, 54 (Supplemental Table 6). The most common issue is exposure classification, which is caused mainly by exposure evaluation at the area level or the use of the nearest monitoring station data. The details of the risk of bias per analysis are presented in Supplemental Table 7.

Prenatal exposure and KD

Three studies evaluated the association between prenatal exposure to at least 1 air pollutant and the risk of KD. Associations were reported for 8 different air pollutants, including PM_2.5, SPM, CO, NO, NO₂, NOx, SO₂, and O₃. Because the number of studies reported for the same air pollutant was insufficient (<3) to support the meta-analysis, only a narrative review was performed. Two associations of NO₂ suggested a positive (but not both significant) correlation with KD. Among the 7 individual associations, 4 pollutants (CO, NO, NOx, and SO₂) had a statistically significant positive correlation with the risk of KD, 2 pollutants (PM_2.5 and SPM) had a nonsignificant positive association with KD, and 1 pollutant (O₃) was significantly negatively associated with KD (Table 2).

Table 2.

The Summary of Pooled Effect Sizes

Pollutant	Number of Studies	Pooled OR (95% CI)a	I2 (%)	Pheter	Pegger	Pbegg
Prenatal exposure
PM2.5	1	2.150 (0.096-42.653)	NA	NA	NA	NA
SPM	1	1.130 (0.790-1.610)	NA	NA	NA	NA
CO	1	1.001 (1.000-1.001)	NA	NA	NA	NA
NO	1	1.024 (1.016-1.032)	NA	NA	NA	NA
NO2	2	2.189 (0.913-5.246)	NA	NA	NA	NA
		1.005 (1.005-1.011)
NOx	1	1.005 (1.005-1.010)	NA	NA	NA	NA
SO2	1	1.489 (1.060-2.052)	NA	NA	NA	NA
O3	1	0.995 (0.990-1.000)	NA	NA	NA	NA
Short-term postnatal exposure
PM2.5	3	1.011 (1.003-1.019)	0	0.97	0.353	0.296
PM10	3	1.004 (1.000-1.008)	38	0.20	0.203	1.000
CO	2	1.000 (0.998-1.003)	NA	NA	NA	NA
		1.000 (1.000-1.001)
NO	1	1.472 (1.004-2.155)	NA	NA	NA	NA
NO2	4	1.003 (0.997-1.008)	46	0.14	0.701	1.000
SO2	4	1.038 (0.942-1.144)	58	0.07	0.213	0.308
O3	3	1.017 (0.965-1.071)	85	<0.01	0.094	1.000
Long-term postnatal exposure
PM2.5	3	1.415 (1.179-1.697)	41	0.19	0.296	0.083
SPM	1	1.110 (0.730-1.680)	NA	NA	NA	NA
CO	1	1.000 (1.000-1.001)	NA	NA	NA	NA
NO	2	1.016 (1.008-1.024)	NA	NA	NA	NA
		0.922 (0.781-1.084)
NO2	2	1.005 (1.000-1.011)	NA	NA	NA	NA
		1.054 (0.948-1.170)
NOx	1	1.005 (1.000-1.005)	NA	NA	NA	NA
SO2	1	1.000 (0.856-1.161)	NA	NA	NA	NA
O3	1	1.000 (0.995-1.005)	NA	NA	NA	NA

CO = carbon monoxide; NA = not applicable; NO = nitric oxide; NO₂ = nitrogen dioxide; NOx = nitrogen oxides; O₃ = ozone; P_begg = significance of the Beggs test; P_egger = significance of the Egger's test; P_heter = significance of the heterogeneity test; PM = particulate matter; SO₂ = sulfur dioxide; SPM = suspended particulate matter.

^aThe pooled OR (95% CI) was calculated by meta-analysis (number of studies ≥3) or presented by the OR (95% CI) of each study (number of studies <3).

Short-term postnatal exposure and KD

Twenty associations between short-term postnatal exposure to at least 1 air pollutant and KD risk were extracted from 7 studies. The associations were reported for 7 different air pollutants, including PM_2.5, PM₁₀, CO, NO, NO₂, SO₂, and O₃. A 10 μg/m³ increase in PM_2.5 and PM₁₀ was associated with a 1.100% (OR: 1.011; 95% CI: 1.003-1.019) and 0.400% (OR: 1.004; 95% CI: 1.000-1.008) increased overall risk of KD, respectively. A positive but nonsignificant association was found between exposure to NO₂, SO₂, or O₃ and KD, with a pooled effect size for a 10 μg/m³ increase in exposure of 1.003 (95% CI: 0.997-1.008), 1.038 (95% CI: 0.942-1.144), and 1.017 (95% CI: 0.965-1.071), respectively. Among the 3 associations identified in the systematic literature review only, 1 for NO indicated a significant positive correlation with KD, and 2 for CO indicated a positive (but not both significant) correlation with KD (Figure 2, Table 2).

Figure 2.

Forest Plot of Short-Term Postnatal Air Pollution Exposure and Kawasaki Disease

Pooled OR and 95% CI for meta-analysis. Population size represents the number of KD cases included in each individual study. Square sizes represent the relative weight of the studies. Study specific estimates are scaled to a standard unit change of 10 μg/m³. HK = Hartung–Knapp; KD = Kawasaki disease; NO₂ = nitrogen dioxide; O₃ = ozone; PM = particulate matter; SO₂ = sulfur dioxide.

Long-term postnatal exposure and KD

Five studies reported 12 associations between long-term postnatal exposure and KD. Associations were reported for 8 different air pollutants, including PM_2.5, SPM, CO, NO, NO₂, NOx, SO₂, and O₃. We found that a 10 μg/m³ increase in PM_2.5 was significantly associated with an increased risk of KD (OR: 1.415; 95% CI: 1.179-1.697). Two articles explored the association between NO and KD, with mixed findings. Two articles observed a positive (but not both significant) association between NO₂ and KD. Among the 5 individual associations, long-term postnatal exposure to CO or NOx was significantly positively associated with KD, whereas SPM, SO₂, and O₃ exposure were not significantly associated with KD. (Figure 3, Table 2).

Figure 3.

Forest Plot of Long-Term Postnatal Particulate Matter With an Aerodynamic Diameter ≤2.5 μm Exposure and Kawasaki Disease

Pooled OR and 95% CI for meta-analysis. Population size represents the number of KD cases included in each individual study. Square sizes represent the relative weight of the studies. Study-specific estimates are scaled to a standard unit change of 10 μg/m³. KD = Kawasaki disease.

Subgroup analyses

The heterogeneity test revealed significant heterogeneity in the included studies for KD risk with short-term SO₂ and O₃ exposure. Therefore, subgroup analyses were conducted to explore the potential causes of heterogeneity.

Supplemental Table 8 shows the results of the subgroup analyses. For the association between short-term postnatal SO₂ exposure and KD risk, adverse effects were significantly greater in the study with smaller population sizes (OR: 1.424; 95% CI: 1.118-1.814). The results for SO₂ across subgroups were otherwise comparable. With respect to the association between short-term postnatal O₃ exposure and KD risk, significant differences between subgroups were found when studies were stratified by the degree of country development and study population size. Remarkably, the number of studies/effect sizes included in these subgroup analyses was small, implying insufficient statistical power for the tests.

Sensitivity analyses and publication bias

Sensitivity analyses and publication bias evaluations were conducted for short-term postnatal exposure to PM_2.5, PM₁₀, NO₂, SO₂, and O_3, as well as for long-term postnatal exposure to PM_2.5. Sensitivity analyses demonstrated the general stability of the main results before and after each study was removed. Notably, removing the Kwon study turned the pooled estimate for PM_2.5 into nonsignificant, and excluding the Lin study turned the pooled estimate for PM₁₀ into nonsignificant. Additionally, the omission of the Fujii study made the pooled estimate for SO₂ significant, and the omission of the Kwon study made the pooled estimate for O₃ significant (Supplemental Table 9). Egger's test and Begg's test revealed no statistically significant publication bias for the above analyses (all P > 0.05).

Quality assessment

Based on the NOS score, all cohort studies were classified as high quality (NOS score: 7-9). Using the modified NOS score, 5 case-crossover studies and 3 time-series studies were classified as high quality (modified NOS score: 4), while 1 time-series study was intermediate quality (modified NOS score: 3) (Supplemental Table 10).

The quality of evidence was high for short-term postnatal exposure to PM_2.5 and PM₁₀ and long-term postnatal exposure to PM_2.5; moderate for short-term postnatal exposure to NO₂ and SO₂; and low for short-term postnatal O₃ exposure. Overall, the lowest-scoring domains were imprecision and inconsistency, whereas the domain that contributed most to the upgrade in evidence quality was the dose‒response trend (Supplemental Table 11).

Discussion

To our knowledge, this is the first systematic review and meta-analysis to comprehensively synthesize the relationship between prenatal and postnatal ambient air pollution exposure and the risk of KD (Central Illustration). We included 13 studies covering 124,857 participants from North America and East Asia. Our meta-analysis revealed significant positive associations between short-term postnatal exposure to PM_2.5 and PM₁₀, as well as long-term postnatal exposure to PM_2.5, and the risk of KD. Additionally, individual studies have demonstrated that gestational exposure to CO, NO, NO_X, and SO₂; short-term postnatal exposure to NO; and long-term postnatal exposure to CO and NO_X could increase the risk of KD. In contrast, gestational O₃ exposure decreased the risk. The probable reason for this is the dose-dependent or threshold effect of prenatal O₃ exposure, but the harmful critical value is still unclear.⁵¹ Further research is also needed to investigate how O₃ affects the development of KD.

Central Illustration.

Prenatal and Postnatal Exposure to Ambient Pollutants Are Associated With Kawasaki Disease

The systematic review and meta-analysis revealed that prenatal and postnatal exposure to several ambient pollutants is associated with an increased risk of KD. The pooled OR (95% CI) was calculated by meta-analysis (number of studies ≥3) or presented by the OR (95% CI) of each study (number of studies <3). CO = carbon monoxide; KD = Kawasaki disease; NO = nitric oxide; NO₂ = nitrogen dioxide; NOx = nitrogen oxides; O₃ = ozone; PM = particulate matter; SO₂ = sulfur dioxide; SPM = suspended particulate matter.

The results of the sensitivity analyses were consistent with the overall results, and all results were not affected by publication bias. To ascertain the sources of heterogeneity, we conducted subgroup analyses stratified by potential factors. The results of subgroup analyses demonstrated that population size appeared to be the source of heterogeneity in the SO₂ group, with significantly larger estimates, and the degree of country development and study population size appeared to be sources of heterogeneity among O₃ studies, with developing countries and smaller study population sizes being more likely to observe KD. Small sample studies are often inferior to large sample studies but tend to have larger effect sizes.⁵⁵ Demographic and epidemiological trends and elevated air pollution levels in middle-/lower-income developing countries amplify the impact of ambient air pollution.⁵⁶

Several potential biological mechanisms underlying the association between ambient air pollution exposure and KD have been suggested. Inflammation, oxidative stress, immune dysregulation, and genetic susceptibility are thought to strongly contribute to KD development.¹⁴

Recent studies have reported that exposure to air pollutants during pregnancy can promote inflammation and oxidative stress, inducing placental inflammation, stress, and immunological changes, leading to adverse outcomes in offspring.^27,57, 58, 59, 60 Maternal systemic inflammation caused by exposure to ambient air pollutants may affect fetal inflammatory and immune responses. Specifically, gestational exposure to air pollutants can affect cytokine secretion and the lymphocyte immunophenotype distribution and increase the C-reactive protein level in human cord blood.61, 62, 63 Moreover, maternal exposure to pollutants during pregnancy may affect fetal lung development, possibly via oxidative stress, which could elevate children's susceptibility to environmental factors throughout life.^64,65 Furthermore, environmental factors, including air pollutants, may trigger epigenetic changes throughout life (which are extremely sensitive during embryogenesis), affecting immune programming and organ development.^64,66

Postnatal exposure to ambient air pollutants, both short-term and long-term, also contributes to the development of KD in children. Air pollutants can generate reactive oxygen species by undergoing redox cycling and depleting endogenous thiols, inducing oxidative stress.^67,68 Multiple cellular sensing mechanisms triggered by pollution constituents, such as toll-like receptors, Nod-like receptors, aryl hydrocarbon receptors, and reactive oxygen species-sensing pathways, are involved in activating proinflammatory cascades.^69,70 Additionally, air pollutants have direct effects on the human immune system, including hyperstimulating innate immunity and affecting adaptive immunity. Air pollutants can stimulate the recruitment and activation of many types of innate immune cells, such as neutrophils and macrophages.²⁰ Moreover, air pollutants can trigger adaptive immunity by enhancing T helper lymphocyte type 2 and 17 responses and can dysregulate antimicrobial immune responses.⁷¹ Furthermore, epigenetic DNA methylation, noncoding RNA regulation, and histone modifications play important roles in mediating the effects of air pollution.⁷² Thus, the hypothesis that exposure to air pollutants can induce oxidative stress, systemic inflammation, immune dysregulation, and epigenetic changes may explain the impact of air pollution on the pathogenesis of KD.

Study strengths and limitations

This is the first comprehensive systematic review of existing epidemiological evidence on ambient air pollution and the risk of KD. We examined a wide spectrum of air pollutants and investigated prenatal and postnatal (both short- and long-term) exposure. We used the Hartung–Knapp approach, which could reduce type I error and provide more robust results, especially when study population sizes differ and the study number is limited.⁷³

Nevertheless, several limitations should be acknowledged. First, most studies focused on short-term postnatal exposure, and research on the effects of prenatal and long-term postnatal exposure was relatively insufficient. More studies on these periods are highly suggested. Second, exposure measurements were inconsistent and not standardized across studies. A considerable proportion rely on distance from monitoring stations, and studies on long-term exposure often fail to consider relocation information, which may lead to certain exposure biases. Studies with improved exposure measurements are needed in the future. Third, we used single-pollutant models instead of multiple-pollutant models because most included studies used single-pollutant models to generate their effect estimates. However, single-pollutant models cannot assess the interactions among air pollutants. Further research using multiple-pollutant models is needed to explore the interactions of concomitant exposure to multiple air pollutants and time-activity patterns. Fourth, the Kwon study contributed the largest weight to our meta-analysis results on short-term postnatal exposure, as further evidenced by the sensitivity analyses. Fifth, although the publication bias tests showed no statistically significant publication bias, the number of studies included for each outcome was below 10 (generally regarded as the minimum threshold for conducting publication bias tests). Sixth, most of the included studies are from Asia. Future research in other regions with high levels of air pollution is needed to increase the generalizability of the findings. Finally, although the included studies adjusted for a range of confounders, confounders differed among these studies, and residual confounders may still be present. These issues limit the interpretability of the findings. Therefore, more detailed consideration of confounders should be taken into consideration.

Ambient air pollution poses a substantial challenge to public health worldwide because it threatens human health, including the probable increased risk of KD. Future studies conducted in other regions with high levels of air pollution investigating prenatal and long-term postnatal exposure and employing standard exposure assessment methods, multiple-pollutant models, and detailed confounders are warranted to provide a complete understanding of the association between air pollution and KD and better risk information to regulatory agencies.

Conclusions

The present systematic review and meta-analysis revealed that prenatal and postnatal exposure to several ambient pollutants are associated with a higher incidence of KD. Appropriate policies and actions are suggested to ameliorate this global public health concern.

Perspectives.

COMPETENCY IN MEDICAL KNOWLEDGE: This systematic review and meta-analysis of 13 epidemiological studies and 124,857 participants revealed that prenatal exposure to CO, NO, NO_X, and SO₂; short-term postnatal exposure to PM_2.5, PM₁₀, and NO; and long-term postnatal exposure to PM_2.5, CO, and NO_X are associated with an increased risk of KD.

TRANSLATIONAL OUTLOOK: Further studies are necessary to explore the potential biological mechanisms underlying the association of prenatal and postnatal ambient air pollution exposure and KD risk and to evaluate whether air pollution prevention and control interventions can reduce KD incidence.

Funding support and author disclosures

The authors have reported that they have no relationships relevant to the contents of this paper to disclose.