Risk Factors for Coronary Artery Aneurysm in a Chinese Pediatric Population with Kawasaki Disease at Low Risk of Intravenous Immunoglobulin Resistance: A Retrospective Cohort Study

Kawasaki disease (KD) is a form of systemic vasculitis of unknown etiology affecting small- and medium-sized arteries that is characterized by fever. With the increasingly common use of intravenous immunoglobulin (IVIG) combined with aspirin to treat KD, the overall case fatality rate has decreased over the years as has the incidence of coronary artery aneurysm (CAA), which has decreased from 25% to approximately 4% [1]. Patients with IVIG resistance are at a considerably increased risk of CAA, which may result in ischemic heart disease, myocardial infarction, or sudden cardiac death at later stages [2‒4]. Given the increased awareness of the severe cardiovascular complications associated with KD, multiple scoring systems for predicting IVIG resistance based on demographic, clinical, and laboratory parameters have been developed [5‒7]. These scores effectively identify patients with KD who have a high risk of IVIG resistance and could benefit from additional anti-inflammatory therapy to reduce their risk of CAA [8‒10]. Patients with KD and low scores are classified as IVIG responders and are characterized as having mild inflammation with a low risk of IVIG resistance [6]. Such classification might lead to the most appropriate management in the acute phase of KD to prevent the development of coronary artery lesions (CALs). However, a universal scoring system suitable for all populations has not been developed due to significant demographic diversity [11‒14]. Stratification strategies using these applied scoring systems were designed to determine which patients would have initial IVIG resistance and not to predict those who would develop CAA. Furthermore, the risk factors for initial IVIG resistance and development of CAA in patients with KD are not similar, with particular differences in infants having been reported [15]. The risk of developing coronary artery disease in patients who respond to IVIG was reported to be between 12.8% and 19.6% [15, 16]. A recent study found that up to 5.5% of Japanese patients with KD and a low risk of IVIG resistance, defined as a Kobayashi score of <5 points, developed CAA [17], indicating that this population should not be assumed to have benign prognoses. Three previous epidemiological surveys from China indicated that the overall trend in CAA morbidity had increased from 5.4% to 20.6% in the past two decades [18‒20]. However, the incidence of CAA among patients with KD and a low risk of IVIG resistance remains unknown since the Kobayashi score does not accurately predict poor outcomes in Chinese populations [13]. Our previous study showed the risk factors for IVIG resistance and CAA development in high-risk patients with KD [21]. Notably, patients with a low risk of IVIG resistance collectively account for most of the patients with KD; therefore, this study focused on the same in patients with KD determined to have a low risk of resistance to initial IVIG treatment, aiming at further CAA incidence reduction.

This study was performed at the Department of Pediatrics, First Affiliated Hospital of Guangxi Medical University, China. The medical records of all children who fulfilled the diagnostic criteria for KD between January 2003 and January 2022 were reviewed. The diagnostic criteria for KD were based on the American Heart Association (AHA) guidelines [4, 22]. A pediatric cardiologist and echocardiographers determined the diagnosis and measurements of CAAs, respectively. CALs were defined according to Z scores adjusted for body surface area (BSA) for a coronary artery with an internal diameter of ≥2.0 in either the proximal right coronary artery (RCA), left main coronary artery (LCA), or left anterior descending artery (LAD); moreover, only cases wherein the maximal Z score persisted at ≥2.5 for more than 1 month after disease onset were considered CAA cases, as described in the AHA guidelines and the guidelines on the diagnosis and management of cardiovascular sequelae in KD (JCS/JSCS 2020) [4, 22, 23]. Exclusion criteria included no IVIG treatment and receiving IVIG or corticosteroid therapy in other medical facilities with incomplete data. Furthermore, patients treated for >10 days after the onset of fever or with recurrent KD were also excluded since the previous CAA could have persisted as a coronary complication. The requirement for informed consent was waived because this was a retrospective data review, and the research was approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Code number: 2021 [KY-E-240]).

All patients received initial IVIG (2 g/kg administered as a single intravenous infusion) with oral aspirin (30–50 mg/kg per day). For patients who were afebrile for more than 3 days, the aspirin dosage was reduced to 3–5 mg/kg per day until patients showed no evidence of CAA 6–8 weeks after fever onset or until CAA regressed. Patients with IVIG resistance were defined as those with symptoms of persistent or recrudescent fever (axillary or rectal temperature: ≥37.5°C and ≥38.0°C, respectively) for at least 36 h and no longer than 7 days after receiving the initial IVIG infusion (2 g/kg) [4]. Patients were grouped retrospectively into the IVIG-responsive (n = 556) and IVIG-resistant (n = 108) groups. Since no universal risk scoring system applies to all populations, 14 existing scoring systems [5‒7, 13, 24‒33] that were previously reported for the determination of patients with KD at high risk of IVIG resistance were applied, and between-group comparisons were performed. Patients were assigned scores according to the rules of each scoring system (see online suppl. File; for all online suppl. material, see https://doi.org/10.1159/000530708). First, eligible patients were diagnosed with low-risk KD using the optimal scoring system, and the association of baseline characteristics with CAA development was investigated in the low-risk group; subsequently, sensitivity analyses were conducted for patients with KD who have a low risk of IVIG resistance as determined by the Kobayashi scoring system, which is widely used for Japanese patients.

A database was prospectively created and retrospectively reviewed. It included (i) general demographic data, including age (in months) at disease onset, sex, and body mass index; (ii) clinical characteristics such as the prevalence of incomplete KD and five principal symptoms, duration of fever before admission, illness day at treatment (illness day 1 = first day of fever), response to IVIG therapy, and total score based on the optimal and Kobayashi scoring systems; and (iii) laboratory indicators, including white blood cell, neutrophil, and lymphocyte counts, hemoglobin concentration, platelet count, C-reactive protein, aspartate aminotransferase, alanine aminotransferase, total bilirubin levels, serum albumin, and serum sodium concentrations. All laboratory indicators were assessed during the acute febrile period and before the initial IVIG treatment. We calculated the neutrophil-to-lymphocyte count ratio (NLR), platelet-to-lymphocyte count ratio (PLR), albumin-to-globulin (A/G) ratio, and C-reactive protein-to-albumin ratio based on these indicators. Echocardiography was routinely performed in the acute stage before IVIG treatment (baseline, days <10) and was repeated 1 month (days 20–40), 2 months (days 50–70), and 3 months (days 80–100) after the onset of fever, and patients were closely monitored depending on the degree of the disease. The Z scores were calculated using the Dallaire equations [34] and plotted against time to evaluate how the size of the CAA evolved.

The normality of the distribution was verified using the Shapiro-Wilk and homogeneity tests. Measurement data with a normal distribution are expressed as means ± standard deviation, and a two-independent sample t test was used to compare such data between the groups. Measurement data without a normal distribution are expressed as medians (4-digit interval) [P₅₀ (P₂₅, P₇₅)], and these data were compared between the groups using the Mann-Whitney U test. Enumeration data are expressed as a percentage (%). The χ² or Pearson’s χ² test was used to perform intergroup comparisons. Variance inflation factors were used to check for collinearity. Factors with a p value <0.1 in the univariable analysis were included in the multivariate logistic regression analysis to determine the risk factors. The optimal threshold for the significant parameter constructed using receiver operating characteristic (ROC) curves and two-tailed p values <0.05 indicated statistical significance. Statistical analyses were performed using SPSS, version 26.0 (IBM Corp., Armonk, NY, USA).

During the study period, 909 children with a diagnosis of KD were admitted. Overall, 245 children were excluded from this study, including 37 children treated without IVIG and 79 treated after the first 10 days of fever onset, 60 who received IVIG or corticosteroid therapy outside the hospital with missing clinical data, 3 who had recurrent KD, and 66 with incomplete clinical or laboratory data. Ultimately, 664 children with KD (108 [16.3%] children with initial IVIG resistance and 556 with IVIG response) participated in this study (Fig. 1); 346 participated in previous studies. The mean age was 30 months (range, 2–108 months) with a male-to-female ratio of 2.3:1 (388 boys and 168 girls) for the IVIG-responsive group and 31 months (range, 1–117 months) with a male-to-female ratio of 2.9:1 (80 boys and 28 girls) for the IVIG-resistant group. Among the IVIG-resistant patients with KD, 38 (5.7%) received additional corticosteroid therapy due to unremitting fever after the end of the second IVIG treatment. First, intravenous methylprednisolone was administered at 2 mg/kg in two divided doses per day. Subsequently, methylprednisolone was tapered and withdrawn until the C-reactive protein level normalized. Next, oral prednisolone was initiated from 2 mg/kg, reduced to 1 mg/kg, and finally to 0.5 mg/kg per day, and tapered for ≥2 weeks. Almost all children were treated with aspirin, and no biological agents, such as infliximab, cyclosporine, anakinra, cyclophosphamide, or plasma exchange, were administered during this period. The routine use of antihistamines before IVIG infusion to avoid hypersensitivity reactions is not recommended in our medical center, and no patient has discontinued treatment due to infusion reactions. Furthermore, no disease-related deaths were observed in either group.

Results of the analysis of the 14 scoring systems by study group are presented in the online supplementary File. All risk scores were higher in the IVIG-resistant group than in the IVIG-responsive group, and the differences were significant (all p < 0.05). The area under the ROC curve (AUC) for the Liping risk scoring system [13] was the largest, with an AUC of 0.714 (95% CI: 0.656–0.771), while the Kobayashi scoring system [6] had an AUC of 0.693 (95% CI: 0.636–0.749) (Fig. 2).

Overall, 444 hospitalized children were diagnosed with low-risk KD according to the optimal scoring system (Liping risk score of <5 points). The incidence of CALs and IVIG resistance in low-risk patients increased with an increasing risk score, and patients with a Liping risk score of 4 points had the highest CAA incidence (26.9%) (Fig. 3). Of the patients included, 100 who developed CAA were enrolled in the CAA sub-group, and the remaining 344 were enrolled in the non-CAA sub-group. The proportion of patients with the following characteristics was higher in the CAA sub-group than in the non-CAA sub-group: age <6 months, male sex, albumin level <35 g/L, and baseline Z score of the coronary artery internal diameter (all p < 0.05). However, no significant differences were observed between the sub-groups for clinical characteristics (all p > 0.05; Table 1).

Univariate analysis revealed six potential risk factors associated with CAA, including age <6 months at fever onset, male sex, albumin level <35 g/L, and the baseline Z score of coronary artery internal diameter. The baseline maximum Z score, rather than the Z scores of the LCA and RCA, was included in the multivariate analysis since there was multicollinearity between the three. Collinearity was absent among these analyzed factors, and age <6 months at fever onset (odds ratio [OR]: 3.274, 95% CI: 1.077–9.950, p = 0.037), male sex (OR: 1.878, 95% CI: 1.001–3.571, p = 0.049), and baseline maximum Z score (OR, 3.437; 95% CI: 2.575–4.589, p < 0.001) were found to be independent risk factors for CAA in patients with KD exhibiting a low risk of IVIG resistance, defined as a Liping score of <5 points. We also adjusted for IVIG resistance, which was previously reported as a risk factor for CAA development [21, 35‒37], and found that an additional adjustment for laboratory indices changed the observed association only minimally; the adjusted odds ratio [aOR] values for age <6 months at fever onset, male sex, and baseline maximum Z score were 3.142, 1.946, and 3.451, respectively (Table 2). Similarly, age <6 months at fever onset (aOR: 3.977, 95% CI: 1.474–10.735, p = 0.006) and baseline maximum Z score (aOR: 3.340, 95% CI: 2.626–4.247, p < 0.001) were also significantly associated with CAA development in 134 (22.6%) patients with CAA among the 594 low-risk patients with KD, defined as a Kobayashi score of <5 points (Table 3).

Multivariate logistic regression analysis showed that a high baseline maximum Z score was significantly associated with CAA; a cutoff value of ≥2.72 yielded a sensitivity and specificity of 96% and 65%, respectively (AUC = 0.844 [95% CI, 0.807‒0.882, p < 0.001]). Overall, 594, 444, and 426 patients were diagnosed using the Kobayashi score, the Liping score, and both scoring systems, respectively. For patients with KD and a Liping risk score of <5 points, the incidence of CAA development was 50.0% (8/16), 26.2% (78/298), and 44.4% (96/216) for age <6 months at fever onset, male sex, and a baseline maximum Z score ≥2.72, respectively. These results were significantly higher than those for age ≥6 months (21.5% [92/428]), female sex (15.1% [22/146]), and baseline maximum Z score <2.72 (1.8% [4/228]) (all p < 0.05, Fig. 4a). Similar results for the incidence of CAA development in patients with KD and a Kobayashi score of <5 points were of particular interest (50% [10 of 20] age <6 months, 21.6% [124 of 574] age ≥6 months, p = 0.007; 25.7% [108 of 420] male sex, 14.9% [26 of 174] female sex, p = 0.004; and 43% [128 of 298] baseline maximum Z score ≥2.72, 2% [6 of 296] baseline maximum Z score <2.72, p < 0.001) (Fig. 4b). Figure 5 shows the incidence of CAA in low-risk patients by risk factor, revealing that the risk of developing CAA increased with the number of risk factors.

Our previous study revealed the risk factors for resistance to IVIG treatment and CAA development in a Chinese pediatric population with high-risk KD. In this study, we further explored the risk factors for developing CAA in a Chinese pediatric population with KD and a low risk of IVIG resistance by expanding the sample size. Overall, 52.1% (346 of 664) of patients participated in our previous studies, and the Liping risk score [13] was observed to have a potential value in predicting IVIG resistance. Among the patients with KD at low risk of IVIG resistance based on a Liping risk score of <5 points, the incidence of CALs and IVIG resistance increased with an increase in the risk score, and patients with a Liping risk score of 4 points had the highest CAA incidence (26.9%). Additionally, the risk factors for developing CAA at 1 month of illness were age <6 months at fever onset, male sex, and a baseline maximum Z score ≥2.72; moreover, the risk increased with an increase in the number of risk factors, with very similar results when applied to patients with KD and at low risk of IVIG resistance as defined by a Kobayashi score of <5 points [6]. To the best of our knowledge, this is the first study to explore the risk factors for developing CAA in a Chinese pediatric population with KD and a low risk of IVIG resistance, providing information for physicians to identify patients at increased risk of adverse outcomes of KD.

KD is a self-limited disease with genetic susceptibility and has a male preponderance. Kitano et al. [38] suggested that the combination of patient age at KD onset, sex, and other predisposing non-modifiable host factors plays a considerable role in developing KD-related CAA. Accumulating evidence [11, 36, 39] has revealed that age <1 year and male sex, which are risk factors for IVIG resistance, are also associated with the development of CAA in patients with KD, which partly aligns with our findings. Our study found that age <6 months and male sex were significantly associated with CAA development. A significant incidence of CAA was observed among sub-groups stratified by these two risk factors. Notably, a similar result was observed for low-risk patients with KD as defined by a Kobayashi score of <5 points. Our findings differed from those of Iio et al. [17], who found that IVIG resistance, rather than male sex, was a risk factor for CAA development in addition to a baseline maximum Z score >2.5 and age <12 months at fever onset. Unexpectedly, a similar incidence of IVIG resistance was observed in patients with and without CAA in our cohort and in patients aged <12 months, probably because of differences in study populations. Compared to previous studies, which included all patients with KD regardless of the Kobayashi score, this study was conducted only on patients with KD at low risk of IVIG resistance based on different scoring systems. Although age <6 months may be a significant risk factor for CAA development in our cohort, among patients with a Kobayashi score of <5 points, IVIG resistance was not a risk factor using either of the scores. Interestingly, the incidence of CAA was significantly different between sub-groups according to age and sex in selected patients when using the different scoring systems, which lends further support to the results of Kitano et al. [38]. The different triggers, weak immune response, and immature coronary arteries of infants may contribute to the vulnerability to CAA formation in infants with KD even if they respond to initial IVIG treatment [40]; however, the real reason remains a mystery. Furthermore, a prospective study on 3,043 people aged 18–30 years revealed that the prevalence of coronary artery calcium was significantly higher in men than in women (15% vs. 5.1%) [41]. As an X-linked gene, CD40 polymorphisms have been repeatedly associated with KD and CAA formation susceptibility in different ethnicities [42‒45], suggesting that genetic factors also contribute to the gender-dependent differences in this condition as well as disease severity. However, the odds of developing CAA in men could not be explained considering the male dominance in KD. Therefore, our results can be used as reference for follow-up management in patients with KD, and further investigation is warranted to better elucidate the combined associations of patient age and sex preponderance with the development of CAA.

Coronary artery internal diameter-based severity is the most significant predictor of a late coronary outcome [46, 47] since the possibility of CAA regression over time may correlate with the initial lesion size [48]. Patients with CALs at the first echocardiography have a higher risk of initial IVIG treatment failure or subsequent progression of CAA [49]. The incidence of CAA (22.6%) in this study was comparable with those reported in the literature (2–24.1%) [22, 36, 50]. However, ours was slightly higher than that in China (20.6%) [20]. Additionally, the CAL prevalence at initial echocardiography (58.1% [258/444] in low-scoring patients based on the Liping risk score and 57.6% [342/594] in patients based on a Kobayashi score of <5 points) was higher than that of the previous study (range, 22–44%) [51‒53]. One reason for this discrepancy may be that the Z score criteria might increase the recognition of CALs. A growing recognition indicates that the criteria for coronary abnormalities, as defined by the Japanese Ministry of Health, are arbitrary, and the more sensitive BSA-adjusted coronary artery Z scores provide some advantages for potential clinical application in predicting the future progression of CALs. Given this background, a high baseline maximum Z score is a risk factor for CAA [15, 54‒56], which conformed to the results of our study and was unchanged even when examining low-risk patients with KD as defined by a Kobayashi score of <5 points. Our previous study revealed that a baseline Z score of LCA ≥2.8 was a risk factor for developing CAA in high-risk patients with KD [21]. A similar trend was observed in patients with KD at low risk of IVIG resistance. A baseline maximum Z score ≥2.72 was strongly associated with CAA development 1 month after disease onset, indicating that the baseline Z score of the coronary artery plays a central role in the likelihood of coronary artery regression within the normal range; this is consistent with the results of Fujiwara et al. [48]. Recent studies indicated that patients with CALs at diagnosis are likely to have CAL progression even after standard treatment [49, 54]. Therefore, the direct observation of the baseline Z score of the coronary artery is important not only to aid in KD diagnosis but also to elucidate in advance the relationship between CAL progression and clinically instructive interventions.

Hypoalbuminemia, the most common type of hepatic dysfunction, is common in patients with KD and is also a variable included in the Harada score [57], which has a cutoff of <35 g/L within 7 days of onset and was designed to predict CAA development. Moreover, hypoalbuminemia is associated with greater severity of illness and progressive coronary dilatation as it is related to increased capillary permeability related to systemic vasculitis [58]. This has been confirmed by the 22nd Japanese epidemiological survey on KD, which indicated that a 1-g/dL decrease in serum albumin levels predicted a 0.66-fold increased risk of coronary dilatation and 0.34-fold increased risk of coronary aneurysm formation [59]. Meanwhile, this study found lower serum albumin levels in the CAA sub-group than in the non-CAA sub-group, with no significant difference; this may be because our study only included patients at low risk of IVIG resistance based on the Liping risk score. Since albumin levels ≤34 g/L was part of this score, it diminished the gap between patients with and without CAA. However, a significant difference in the proportion of those with an albumin level <35 g/L was observed between both groups. The serum albumin level was negatively associated with CAA development in patients with high-risk KD in our previous study [21], suggesting that hypoalbuminemia is a characteristic of patients with CAA.

This study has some limitations that should be addressed. First, the retrospective nature of this study has inherent limitations. Second, the collection of limited laboratory data precluded testing for other scoring systems, and the study population was entirely composed of Chinese patients. Therefore, our results may not apply to other ethnicities. However, the results obtained for patients with KD and a low risk of IVIG resistance, as defined by the Kobayashi score, were similar. Third, the diameter of the LAD, particularly in patients without CAA, could only be obtained in a small population and, therefore, could not be analyzed. Furthermore, the quality of echocardiography may have varied due to differences in operator expertise, machine quality, and choice of sedation, leading to an overestimation of the true incidence of CAA. Therefore, a multi-center study with a large cohort is required.

To further prevent CAA development in patients with KD, we can consider predicting IVIG responders based on the CAA risk factors identified in this study as candidates for future clinical trials with intensified primary IVIG treatment and adjunctive therapies; however, close follow-up should be considered. Additionally, future multi-center-based studies are needed to validate the predictive value of the identified risk factors in different ethnic populations.

We thank Dr. Cheng Chen for the helpful advice and discussions. We also gratefully acknowledge Yuqin Huang, Qiaoyu Que, and Kaizhi Liang from the Pediatrics Department of First Affiliated Hospital of Guangxi Medical University who instructed and assisted us during the manuscript preparation.

This study was approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Code number: 2021 [KY-E-240]). Patient consent was waived due to the retrospective nature of this study.

The authors have no conflicts of interest to declare.

Guangxi Medical and Health Key Discipline Construction Project 2019 [19], and Guangxi Clinical Research Center for Pediatric Disease (No.: GUI KE AD22035219).

Jie Liu drafted the manuscript, contributed to data collection, performed statistical analysis, and approved the final manuscript as submitted. Danyan Su, Piaoliu Yuan, and Bingbing Ye provided the figures, contributed to data collection and study design, and approved the final manuscript as submitted. Suyuan Qin administered primary treatment to the patients while they were admitted, contributed to the study design, and approved the final manuscript as submitted. Yusheng Pang conceived and designed the study, contributed to data collection, and approved the final manuscript as submitted.

All data generated or analyzed during this study are included in this article and its online supplementary material. Further enquiries can be directed to the corresponding author.