We performed an electronic systematic literature search on the platforms of PubMed, Embase, and the Cochrane library for studies on KD, from inception to 24 July 2021. In the current study, three well-trained authors (L.S. Chang, P.C. Chen, and C.W. Hsu) independently performed a systematic literature search from the study's inception and scanned the outcomes of the search based on the abstracts and titles using a predetermined list of variables of interest.

The search keywords were “Kawasaki disease^*” AND (allerg^* OR atop^* OR asthma OR rhinitis OR dermatitis OR eczema) and were adjusted according to each database's requirements (17). The review questions were formulated following the PICO criteria (Population, Intervention, Comparator, and Outcome) in the Embase database using the keywords: “mucocutaneous lymph node syndrome” /exp AND (“allergy” /exp OR “allergic condition” OR “allergic state” OR “allergy” OR “allergy, physical” OR “physical allergy”). The present study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) (Figure 1; Supplemental Table 1) (18). Furthermore, manual searches were performed for potential papers selected from the reference lists of the included articles to expand our eligible list.

We included both case-control and cross-sectional designs that were conducted in patients after KD diagnosis. Regarding the outcome and exposure measures, any assessment method of KD, asthma, allergic rhinitis, and/or atopic dermatitis was included in the narrative review, as well as in the quantitative meta-analysis. We imposed no date or types of publication restrictions on the search. The exclusion criteria were studies that reported allergic diseases before or concomitant with KD development, studies without distinguishable comparison groups, or articles that did not provide original data or full-texts. We also excluded studies not written in English. An overlapping or similar dataset used in different published articles or a dataset using a cohort study was excluded.

During the screening stage, the three authors screened all of the titles and abstracts of the identified results. Studies not relevant or available to the prevalence or incident rate of allergic diseases including asthma, allergic rhinitis, or atopic dermatitis in patients after KD were excluded in this stage. We subsequently performed examination of full-texts after developing a list of potentially eligible studies.

Between-reviewer inconsistencies concerning eligibility, inclusion, or data extraction were settled through discussion after reading the full text under the supervision of the corresponding authors followed by mutual consensus. We were thus able to determine a final list of included studies.

Either crude data on exposure and outcome or the odds ratio (OR) of the association between an allergic disease and KD needed to be recorded. The independently extracted data of interest comprised information on the study design and characteristics, including the first author, published year, number of participants, sex ratio in the form of the percentage of males, mean age, crude unadjusted study outcome (two by two contingency table), study outcome adjusted for confounding factors (adjusted OR), information on confounders, and the assessment method of both the allergic diseases and KD.

We set the primary outcomes as the OR of the association between asthma, allergic rhinitis, or atopic dermatitis following KD. We aimed to investigate studies that examined the prevalence and/or incidence of one or more of the three allergic diseases, including asthma, allergic rhinitis, and atopic dermatitis, in subjects with KD compared with those without KD. Data was calculated from the information available in the manuscripts without contacting the original authors.

We investigated the association of KD with the development of the three allergic diseases using Comprehensive Meta-Analysis software 3.3 (CMA, Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. Biostat, Englewood, NJ 2013). We adopted CMA to determine the potential influences of induvial studies by sensitivity analysis of one-study removement. Subgroup analyses were conducted to determine the difference between research methods. Based on the assumed heterogeneous background and populations among the recruited studies, crude data were pooled using the random effect model meta-analyses to determine the OR with the 95% confidence interval (CI).

I² tests were assessed for statistical heterogeneity in the included studies. Potential publication bias was simultaneously evaluated by visual inspection of funnel plots and Egger's regression test (19).

We used the modified version of the Newcastle-Ottawa Scale (NOS; Stang, 2010) designed to assess the quality of observational studies, ranging from zero (low quality) to ten (high quality) for case-control and cross-sectional studies to determine the methodological quality of the included studies (19–21). No threshold of NOS has been validated. According to a previous meta-analysis, we defined studies as being low quality (score <5), moderate quality (score 5–6), or high quality (score >6) (22).

In this review, the total yield was 1,038 articles, of which 229 duplicate studies were removed. We then excluded 799 studies based on their titles and abstracts, leaving 10 records to be retrieved for further screening. Only four studies were not based on the Taiwan Health Insurance Database, two of which were cross-sectional studies, and the other two were case-control studies (23–26). The majority of cross-sectional and longitudinal studies reported a statistically significant positive association. On reviewing full-text articles, we excluded two cohort studies duplicated with the research of the Taiwan Health Insurance Database, one that included only admission population, and three studies that included participants with allergic diseases before or concomitant with KD onset (4, 7, 8, 13, 16, 26). Four studies in four countries representing 8,474 participants were finally included in the meta-analysis (23–25, 27). The proposed algorithms with the number of studies are illustrated in Figure 1.

The main characteristics and population of the four studies included in the meta-analysis are depicted in Table 1. Regarding types of allergic diseases assessed among patients with KD, one study included only the outcome of asthma and allergic rhinitis (23). CMA was performed for the results of asthma, allergic rhinitis, and atopic dermatitis after KD. The studies included in this systematic review were published between 1997 and 2016, and the sample size of KD patients ranged from 9 3 in Singapore to 7,072 in Taiwan, including 5,304 males and 3,170 females. More than half of the participants were males (62.59% in KD children and 58.58% in controls). The average number and standard error of subjects was 2,118.50 ± 1,669.52. Of the four studies included in this review, one was from Israel, one from Singapore, one from Japan, and one from Taiwan, with varied data sources. In this analysis, the vast majority were Asians, including Chinese, Indians, Japanese, and Malays (24, 25, 27). Studies used such assessment methods as interview via telephone, questionnaire (parental report or ISAAC, i.e., the International Study of Asthma and Allergies in Childhood), code in database, or medical records. The number of the studies and sample size for each allergic disease were as follows: asthma, four studies, n = 8,474; allergic rhinitis, four studies, n = 8,474; atopic dermatitis, three studies, n = 8,330.

Four studies that reported asthma among patients with KD were included in the final analysis (Table 1). Of the 8,474 children with Kawasaki disease, 773 were diagnosed with asthma. Based on the results of the random-effects meta-analysis model, the pooled estimated OR of children having KD compared to those without was 1.437 (95% CI 1.067–1.937), demonstrating a statistically significant difference (p = 0.017) and indicating an association between KD and asthma. We found medium heterogeneity for this analysis (I² = 74%; p = 0.009) (see Figure 2A). To explore the influence of case-control studies vs. cross-sectional studies on the effect estimate, we conducted subgroup analysis. The result from the subgroup analysis showed that a higher proportion of children in the cross-sectional design subgroup (two studies, OR: 2.552, 95% CI: 1.588–4.102) had asthma compared to the children in the case-control subgroup (two studies, OR: 1.164, 95% CI: 1.069–1.267, p = 0.001).

Four studies examined the risk of allergic rhinitis in children with KD compared to children without KD. The meta-analysis showed that children with KD were likely to have allergic rhinitis compared with those in the control groups (OR: 1.726; 95% CI; 1.291–2.307). We found an apparent heterogeneity among the studies included in this meta-analysis (I² = 89%; p < 0.001) (see Figure 2B). Regarding subgroup analysis, no apparent difference was detected between the case-control and cross-sectional studies (p = 0.070).

The summary OR of atopic dermatitis among KD patients was 1.243 (95% CI: 0.857–1.802; three studies, n = 8,330 KD children) (Figure 2C). This finding showed that, while children with KD have more atopic dermatitis compared with those in the control groups, it does not reach statistical significance. The heterogeneity across the studies was found to be statistically significant (I² = 89%; p < 0.001).

In the meta-analysis of the asthma group, removing studies performed by Matsuoka et al. or Tsai et al. affected the asthma risk following KD (Supplemental Figure 1A) (25, 27). Both research groups contain more than 1,000 KD children. Removing any study did not affect the significance in risk of allergic rhinitis following KD (Supplemental Figure 1B). In other words, the significance was not impacted by only one study. In the meta-analysis of the effect of KD on the subsequent risk of atopic dermatitis, removing the study by Tsai et al. increased the risk of atopic dermatitis following KD (Supplemental Figure 1C) (27). Considering that the ages of atopic dermatitis and KD are similar or even younger, if we replace the research in the Taiwan Health Insurance Database with atopic dermatitis comorbid with KD, the main results of meta-analysis change into significantly higher co-morbidity rates of AD in subjects with KD from OR: 1.243, 95% CI: 0.857–1.802 to OR: 1.555, 95% CI: 1.175–2.057 (13). The meta-analysis showed that children experienced 55% greater odds of having AD before, after, or along with KD compared with those without KD.

Across the four studies, the average NOS for quality assessment was 7, as shown in Supplemental Table 2. Of the included studies, two studies were of high quality with NOS score 7 and 9, and two were of moderate quality with NOS score 6. Egger's tests did not show any risk of publication bias (p = 0.183 for asthma, p = 0.203 for allergic rhinitis, p = 0.991 for atopic dermatitis) in Supplemental Figure 2.