Is there an association between intravenous immunoglobulin resistance and coronary artery lesion in Kawasaki disease?—Current evidence based on a meta-analysis

Background Coronary artery lesion (CAL) caused by Kawasaki disease (KD) is a leading cause of acquired heart disease in children. Initial treatment of intravenous immunoglobulin (IVIG) can reduce the incidence of CAL. Although most of the current studies have shown a certain correlation between CAL and IVIG resistance, the conclusions are not completely consistent. Thus, we performed this meta-analysis to evaluate the association between IVIG resistance and CAL in KD. Methods PubMed, EMBASE, the Cochrane Central Register of Controlled Trials, and China National Knowledge Infrastructure through April 21, 2020 were searched to detect relevant studies. Data analysis was performed with STATA 15.1. Results A total of 53 relevant studies were eligible to this analysis, including 30312 KD patients, of which 4750 were IVIG resistance and 25562 were responders. There was a significant difference found between IVIG resistance and IVIG response groups in the incidence of CAL (P < 0.001, odds ratio (OR), 3.89; 95% confidence interval (CI) (3.18, 4.75)). The heterogeneity test results showed that the I2 value was 74.8%. The meta-regression analysis showed that the study regions might be the sources of heterogeneity. The subgroup analysis suggested that the incidence of CAL in the IVIG resistance group was still higher than that in the IVIG response group under different regions, IVIG resistance diagnostic criteria, CAL diagnostic criteria, and study types. Meanwhile, the sensitivity analysis did not find any significant impact from every single study. Conclusions This is the first meta-analysis to reveal the incidence of CAL was associated with IVIG resistance in KD patients. Further well-designed studies with uniform criteria are needed to evaluate the incidence of CAL in IVIG resistant patients.


Introduction
Kawasaki disease (KD) is an acute vasculitis of unknown etiology that predominantly affects children, first identified in Japan and now reported worldwide [1]. KD may cause coronary artery lesion (CAL) and is currently the leading cause of acquired heart disease in children in developed countries [2]. The American Heart Association (AHA) recommends the standard treatment regimen for the acute phase of KD involves administering intravenous immunoglobulin (IVIG) 2 g�kg-1 and aspirin [3]. Previous studies have indicated that the incidence of CAL is highly correlated with dose and infusion timing of IVIG, not aspirin [4,5]. Early use of IVIG in KD can effectively reduce the incidence of CAL from 25% to about 4% [3]. However, up to 20% of KD patients may fail to respond to IVIG [6]. Even though the precise molecular mechanism of CAL and IVIG in KD are still unclear, many clinical studies suggest IVIG resistance has deeply associated with the occurrence of CAL [7][8][9]. Thus, a great deal of literature have used this as a starting point to study the indicators for predicting IVIG resistance [10][11][12][13], with the expectation that additional therapeutic measures can be taken to reduce the incidence of CAL through early diagnosis of IVIG resistance.
Although most of the current studies have shown the association between CAL and IVIG resistance, the conclusions are not completely consistent. Besides, there is still a lack of comprehensive and systematic analysis of this issue. Therefore, we performed this meta-analysis for the first time to evaluate the association between IVIG resistance and CAL in KD.

Study protocol
We generated this meta-analysis followed a predetermined protocol by the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [14]. The PRISMA checklist could be found in S1 Table. This study was registered with PROSPERO (CRD42020181359).

Study selection
We initially screened researches by the title and abstract. A full-text search that retrieved potentially relevant reports was assessed for compliance using the inclusion and exclusion criteria. Inter-rater reliability for the study selection was calculated using the kappa statistic [15]. Studies that meet all the following criteria were included: 1) All cases were KD patients; 2) non-randomized or randomized controlled clinical trials or cohort studies evaluating the incidence of CAL in KD patients, including IVIG resistance and IVIG response. Meanwhile, researches meeting any of the following criteria were excluded: 1) Initial IVIG treatment is not the AHA recommended standard dose (2 g�kg -1 ); 2) conference articles, reviews, or abstracts; 3) The sample size included in the study was too small (n < 60); 4) duplicate reports.

Data collection and assessment of study quality
Two investigators (Xiaolan Zheng, Jinhui Li) independently assessed the eligibility of studies by the title and abstract, and a third reviewer (Peng Yue) determining divergence based on the inclusion or exclusion criteria and the quality of literature, and consultation with a fourth reviewer (Yifei Li) if necessary. Besides, we used the Newcastle-Ottawa Scale (NOS) to assess the quality of the included researches [16]. Additionally, literature with a score of fewer than five stars was considered as of low quality and needed to be excluded.

Data synthesis and analysis
We calculated the odds ratio (OR) with 95% confidence intervals (CI) for each outcome of interest then pooled them into an independent meta-analysis. KD children with IVIG resistance were the experimental group, and the control group was the IVIG responders. The outcome was the incidence of CAL in the experimental group or the control group, including coronary artery aneurysm (CAA), coronary artery dilatation (CAD) or ectasia, which was defined according to the diagnostic criteria of the American Heart Association (AHA) Z-score [3] or the Japan Kawasaki Disease Research Committee (JKDRC) [17].

Publication bias
We used Egger's regression by STATA software version 15.1 to test the publication bias of the include studies. Each study is represented by a dot, and the asymmetrical distribution of dots between the regression line suggested publication bias, where the quantified result showed that P < 0.05 [18]. When publication bias occurs, we use the trim-and-fill method further to evaluate the impact of publication bias on the results. If there is no significant difference between the results before and after the test of the trim-and-fill method, it indicates that the results are stable and reliable even if there is publication bias [19].

Heterogeneity
Heterogeneity was investigated by the Chi-squared test and was considered as statistically significant when P < 0.05 in these tests. The I 2 values can range from 0 to 100%, higher than 25%, 50%, and 75%, suggesting low, moderate, and high heterogeneity, respectively [20]. When the I 2 value exceeded 50%, we used the random-effects (M-H method) for analysis; otherwise, the fixed-effects (M-H method) was used.
The meta-regression was conducted to investigate the potential factor of leaded heterogeneity. The subgroups analysis was performed according to the study regions, IVIG resistance diagnostic criteria, CAL diagnostic criteria, and study types.

Sensitivity analysis
Sensitivity analysis was conducted for every study to investigate whether any single study caused undue weight in the results of the meta-analysis.

Statistical analysis
Data analysis was performed with STATA 15.1. The numbers of children who developed CAL in each study were converted to OR. Meanwhile, heterogeneity, publication bias, and sensitivity analysis were also conducted by STATA 15.1.

Search results
Initially, 1232 potentially relevant studies were retrieved by the search method, of which 126 were considered as of interest by titles and abstracts. However, 73 studies were excluded by reading the complete researches due to article type (n = 18), using the modified IVIG dosing as the initial treatment (n = 7), and lacking specific numbers of CAL cases in KD patients with IVIG resistance or response (n = 48). Ultimately, 53 articles [8,10,11,13, were included in the meta-analysis. The kappa test showed the kappa value of agreement during the systematic searches was 0.86. The research selection process was shown in Fig 1.

Study characteristics
A total of 30312 KD patients were included, of which 4750 were IVIG resistance, and 25562 were responders. The 53 studies included were from 11 countries in three regions (Asia, North America, and Europe). Besides, the criteria for the diagnosis of KD resistance was not consistent, among them, IVIG resistance of six studies was defined as persistent fever � 24h after the end of the first IVIG infusion, 21 studies to define the time for � 36h, or otherwise, 17 studies was defined as � 48h, and for the remaining nine studies was defined as a second dose of IVIG for persistent or recrudescent fever without mentioning the specific time. Moreover, the CAL diagnostic criteria in all the included literatures were not consistent, among which, JKDRC criteria was used in 29 studies, AHA criteria was used in 14 studies, JKDRC or AHA criteria was both used in eight studies, and the remaining two studies did not report which criteria was used. All included studies were non-randomized trials, including 5 prospective studies and 48 retrospective studies. Moreover, all the included reports were awarded � six stars and qualified as high quality. The characteristics of KD children in all involved studies were illustrated in Table 1.

Data synthesis and analysis
The OR value and 95%CI of the CAL cases of IVIG resistant children in each study were calculated by STATA 15.1, and all the results were pooled into an independent meta-analysis. The results showed a significant difference between IVIG resistance and IVIG response groups in the incidence of CAL (P < 0.001, OR, 3.89; 95%CI (3.18, 4.75)) (Fig 2). Due to the I 2 value was 74.8%, we used the random-effects (M-H method) for analysis.

Publication bias
The result of the Egger's regression test was shown an asymmetric distribution indicated that publication bias existed (P = 0.002, t = 3.19, 95%CI (0.47, 2.07)) (S1A Fig). Then, we use the trim-and-fill method further to evaluate the impact of publication bias on the results. The funnel plot of publication bias used the trim-and-fill method showed that no significant difference between the results before (P <0.001, Z = 13.26, 95%CI (1.16, 1.56)) and after filled 13 studies

Meta-regression and subgroup analysis
As the results of meta-analysis showed moderate to high heterogeneity, we used the metaregression to investigate the origins of heterogeneities. According to the results (Fig 3), the study regions might be the sources of heterogeneity (P = 0.012, t = -2.62, 95%CI (0.42, 0.89)) ( Fig 3A). Besides, IVIG resistance diagnostic criteria were not the impact factor on the heterogeneity (P = 0.269, t = -1.12, 95%CI (0.63, 1.14)) ( Fig 3B). Meanwhile, CAL diagnostic criteria could not impact the homogeneity too (P = 0.473, t = -0.72, 95%CI (0.65, 1.22)) ( Fig 3C). Additionally, the meta-regression also suggested the study type was not a dramatic impact factor (P = 0.571, t = 0.57, 95%CI (0.52, 3.18)) ( Fig 3D). The subgroup analysis suggested that the incidence of CAL in the IVIG resistance group was still higher than that in the IVIG

Sensitivity analysis
Sensitivity analysis was performed by STATA 15.1, and we did not detect any significant impact from every single research in the result (Fig 4).

PLOS ONE
Is there an association between IVIG and CAL in KD?

Discussion
CAL caused by KD is a significant component of acquired heart disease in children in many countries [70,71]. Although initial IVIG treatment can effectively reduce the incidence of

PLOS ONE
Is there an association between IVIG and CAL in KD?
CAL, IVIG resistance still exists in some children. This is the first meta-analysis to evaluate the association between IVIG resistance and CAL in KD. A total of 53 relevant studies were involved in our meta-analysis, including 30312 KD patients (4750 IVIG resistant patients and 25562 responders). The results showed a significant difference between IVIG resistance and IVIG response groups in the incidence of CAL in KD patients. CAL is the major complications to KD, which reveals a poor prognosis. So that, hundreds of researchers had been involved in the studies of risk factors related with CAL in the last decades. IVIG resistance is one of the dominant adverse effects during acute phase treatment, which brought into observations. Even several previous studies [7][8][9] have suggested that the incidence of CAL was significantly higher in IVIG resistant populations. However, this issue still be debating as a series of researches demonstrated the negative association between IVIG resistance and CAL. So that it is important to draw a clear relationship of IVIG administration effects and CAL based on this meta-analysis. To predict IVIG resistance, a variety of scoring systems have been designed, and Kobayashi and Egami are the most popular ones [72,73]. Additionally, a lot of researches have been conducted to identify risk factors for IVIG resistant KD [74][75][76][77][78]. Patients' ages, biomarkers including CRP, ESR, total bilirubin, AST, ALT, and Pro-BNP were identified positively correlation with IVIG resistance. Unfortunately, the sensitivity and specificity of such indicators are not qualified to be applied alone. So that an integrative scoring system need to be developed to predict IVIG resistance. And the different genetic backgrounds were also considered to be involved in various clinical results. Therefore, our study was intended to conduct subgroup analysis on ethical race. Due to the included literature was mainly divided by region and lacked ethnic information, this paper finally adopted region for subgroup analysis. In addition, some studies have shown that children with younger or elder age are more likely to demonstrate a drug resistance. However, since the included literature in this paper also lacks the segmentation of the age of the included children, further subgroup analysis at different ages cannot be carried out. These questions need to be analyzed in more well-designed studies in the future.
It is worth noting that in our literature search process, we found that majority of studies indicated that IVIG resistance might lead to an increase in the incidence of CAL, while there were few studies with different opinions. In this case, we should be aware of the possibility of publication bias. And then, the Egger's test turned out publication bias existed. Next, we use the trim-and-fill method to evaluate the impact of publication bias on the results. Finally, the trim-and-fill method suggested the results were relatively robust even with publication bias.
As the heterogeneity test results showed that the I 2 value was 74.8%, we conducted the meta-regression analysis to investigate the origins of heterogeneities. The results showed that the study regions might be the sources of heterogeneity. The subgroup analysis suggested that the incidence of CAL in the IVIG resistance group was still higher than that in the IVIG response group under different regions, IVIG resistance diagnostic criteria, CAL diagnostic criteria, and study types. These results indicate that the IVIG resistance group has a higher risk of CAL than the IVIG response group, even under a variety of conditions. Since IVIG was first used in Kawasaki disease in 1982 [79], CAL has decreased significantly. Gradually, the occurrence of IVIG resistance has attracted extensive attention. Research on the prediction and treatment of IVIG resistance has been a research hotspot for many years. Nevertheless, the mechanism by which IVIG reduces the incidence of CAL and IVIG resistance is not yet clear, and the mechanism by which CAL is also unclear, probably because the etiology of KD is still unknown.
Current follow up drug therapy for KD children with IVIG resistance mainly includes the second dose of IVIG, methylprednisolone (MP), infliximab, interleukin 1 receptor antagonist or cyclosporin [3]. However, KD patients who fail to respond to 2 doses of IVIG present a unique challenge [80]. A meta-analysis comprising 372 refractory KD patients conducted in 2019 by Chan et al. [81] revealed that infliximab, MP, and second IVIG infusion showed no significant differences in the cardioprotective effect or the rate of treatment resistance. So far, there is still no accepted treatment algorithm for refractory KD patients [82]. Among the 53 literatures included in our study, there were few literatures that provided detailed adjunct therapies and outcomes for IVIG resistance, so we could not perform analysis to evaluate the effect of different treatment regimens on the incidence of CAL in IVIG resistance. A meta-analysis [83] focused on this issue showed that steroids were more effective at reducing fever than second dose of IVIG, but was no difference in the incidence of CAL. It should be noted that the current AHA guidelines do not recommend the routine use of steroids [3]. Besides, studies which comparing infliximab with the second dose of IVIG in IVIG resistance indicated that infliximab reduced fever duration, but the outcome of coronary was similar [84,85]. In general, there have been no robust clinical trials comparing second-line regiments for IVIG resistance in KD [86], further research into adjunct therapies is an area for future work.
Our study has several limitations. First, most studies lacked information about the adjunct therapies of IVIG resistance. Although previous studies suggested that adjuvant therapy had no significant effect on the incidence of CAL in IVIG resistance, it may still cause bias to the results of this study. Second, the follow-up time of many studies was not reported, which may also lead bias to the results. Third, most of the included studies were retrospective trials, which could lead to biased results. Therefore, further well-designed studies with uniform criteria are needed to evaluate the incidence of CAL in IVIG resistant patients and IVIG responders.

Conclusions
This is the first meta-analysis to reveal the incidence of CAL was associated with IVIG resistance in KD patients. Further well-designed studies with uniform criteria are needed to evaluate the incidence of CAL in IVIG resistant patients.
Supporting information S1 Table.