We conducted a retrospective cohort study including 529 patients with KD who were admitted to the Pediatric Department, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, between November 2019 and December 2024. The diagnosis of KD was established according to the American Heart Association (AHA) criteria (23). The detailed process of patient screening and selection is illustrated in Figure 1. All clinical data and biological samples were collected before December 2024, while data cleaning and statistical analyses were completed by April 2025.

The study protocol was reviewed and approved by the Ethics Committee of Xinhua Hospital, Shanghai Jiao Tong University School of Medicine (Approval No. XHEC-D-2025-093).

The diagnosis of classic KD is based on the presence of ≥5 days of fever (first calendar day of fever is illness day 1) and the presence of ≥4 of the 5 principal clinical features (23): 1) erythema and cracking of lips, strawberry tongue, and/or erythema of oral and pharyngeal mucosa; 2) bilateral bulbar conjunctival injection without exudate; 3) rash: maculopapular, diffuse erythroderma, or erythema multiforme-like; 4) erythema and edema of the hands and feet in acute phase and/or periungual desquamation in subacute phase; 5) Cervical lymphadenopathy (≥1.5 cm diameter), usually unilateral.

The diagnosis of incomplete (sometimes referred to as atypical) KD should be considered in any infant or child with prolonged unexplained fever, fewer than 4 of the principal clinical findings, and compatible laboratory or echocardiographic findings (23).

Children were classified into two groups: IVIG-responsive KD, defined as defervescence without recurrence after the initial IVIG infusion, and IVIG-resistant KD, defined as persistent or recurrent fever (≥38 °C) occurring at least 36 hours after completion of the first IVIG infusion (23).

Children were excluded if they met any of the following criteria: (1) Children with missing data during treatment; (2) Children diagnosed with infection upon enrolment; (3) Children who were given non-standard IVIG treatment; (4) children who had received IVIG or corticosteroid treatment prior to treatment; (5) children with serious complications outside the cardiovascular system; (6)children were readmitted for Kawasaki disease; (7)children with severe immunological disorders.

During the acute phase, IVIG was administered to all enrolled patients at a total dose of 2 g/kg, combined with oral aspirin at 30–50 mg/kg/day. After becoming afebrile for the period of 48 to 72 hours, the daily dosage of aspirin was reduced to 3–5 mg/kg. Patients who showed signs of IVIG resistance were given methylprednisolone (2 mg/kg/day) and an extra IVIG infusion (single dose of 2 g/kg).

All enrolled patients’ medical records were retrospectively reviewed, including demographic data such as age (months), gender, weight, days of fever before admission, clinical characteristics, IVIG response, and laboratory parameters before IVIG therapy.

C-reactive protein (CRP), white blood cell count (WBC), Neutrophil%, Lymphocyte%, absolute lymphocyte count (ALC), hemoglobin (HB), platelet count (PLT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), sodium (Na), total protein (TP), albumin (ALB), interleukin-8 (IL-8), interleukin-1B (IL-1B), interleukin-6 (IL-6), IL-10, tumor necrosis factor-α (TNF-α), and IL-2R were the laboratory parameters.

Serum cytokine concentrations (IL-6, IL-8, IL-10, IL-1β, TNF-α, and IL-2R) were determined using an electrochemiluminescence (ECL) immunoassay on the Cobas e 601 analyzer (Roche Diagnostics, Basel, Switzerland). Venous blood samples for cytokine measurement were collected at the time of hospital admission, before administration of any medication (including IVIG and aspirin), and during febrile episodes (≥38.0 C).

Categorical variables were expressed as counts and percentages, and group differences were compared using the χ² test. Continuous variables were presented as the mean ± standard deviation or median (25th–75th percentile), and independent sample T test or Mann–Whitney U test was used to compare the differences between the two groups depending on the data distribution. Variables with p < 0.05 in the univariate analysis were further included in a multivariate logistic regression analysis to identify independent risk factors for IVIG-resistant KD. Multivariable logistic regression models were constructed, with adjustments made for age, sex, and weight as potential confounding variables. In addition, subgroup analyses were performed by clinically established age groups (e.g., <12 months, 12–60 months, >60 months). The variance inflation factor (VIF) was used to assess multicollinearity, and logistic regression (univariate and multivariate) analyses were performed to identify risk variables. The predictive performance of each biomarker was assessed using receiver operating characteristic (ROC) curve analysis, and the area under the curve (AUC) was calculated. Optimal cutoff values were determined based on the Youden index. All statistical tests were two-sided, and a p < 0.05 was considered statistically significant. Analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA).

Between November 2019 and December 2024, a total of 646 KD patients had been treated at the pediatric department of our institution. Based on the exclusion criteria, 117 cases were excluded. Our analysis ultimately comprised 529 KD patients (Figure 1), 88 of whom (16.6%) were identified as IVIG-resistant.

Baseline demographic and clinical characteristics of the IVIG-responsive (55.1% male) and IVIG-resistant groups (60.2% male) are summarized in Table 1. Compared with the responsive group, IVIG-resistant patients showed a significantly higher incidence of CALs (p < 0.05). Laboratory values are presented in Table 2. IVIG-resistant patients showed significantly lower levels of ALC, PLT, TP, ALB and Na, while CRP, Neutrophil%, ALT, IL-6, IL-8, IL-10, TNF-α, and IL-2R levels were significantly elevated (all p < 0.05).

To identify independent risk factors for IVIG-resistant KD, fourteen variables (ALC, PLT, TP, ALB, Na, CRP, Neutrophil%, ALT, IL-6, IL-8, IL-10, TNF-α, IL-2R and CALs levels) were included in a multivariate logistic regression analysis. After adjusting for confounding factors such as age, gender, and weight, the multivariate analysis indicated that CALs, IL-10, and IL-2R were significant independent predictors of IVIG-resistant KD (p<0.05). Multicollinearity analysis confirmed that the variance inflation factor (VIF) for all variables was below 5, indicating no significant multicollinearity among the variables (Supplementary Table S1). The results showed that CALs, IL-10 and IL-2R levels were independent predictors of IVIG resistance in children with KD (Table 3).

We performed multivariate logistic regression analysis to assess the association of IL-10 and IL-2R levels with IVIG resistance across different age subgroups. Subgroup analysis revealed that IL-10 was an independent risk factor for IVIG resistance across all age groups: < 12 months (OR = 1.169, 95% CI: 1.002-1.363, P = 0.0466), 12–60 months (OR = 1.333, 95% CI: 1.208-1.471, P < 0.0001), and > 60 months (OR = 1.415, 95% CI: 1.164-1.719, P = 0.0005). IL-2R served as a significant risk factor for IVIG resistance in the 12–60 months (OR = 1.121, 95% CI: 1.087-1.155, P < 0.0001) and > 60 months (OR = 1.106, 95% CI: 1.055-1.159, P < 0.0001) age groups but was not statistically significant in the < 12 months age group (OR = 1.026, 95% CI: 0.989-1.065, P = 0.1660) (Table 4).

ROC analysis was used to evaluate the predictive values of IL-10, IL-2R, and the combination of IL-10 and IL-2R for IVIG resistance in children with KD. ROC analysis of the entire cohort showed that IL-10 alone had limited predictive value (AUC = 0.767, sensitivity 78.23%, specificity 61.36%), while IL-2R demonstrated high discriminative power (AUC = 0.825, sensitivity 70.98%, specificity 85.23%). The combination of IL-10 and IL-2R slightly enhanced the predictive efficiency, leading to a balanced improvement in sensitivity while maintaining good specificity (AUC = 0.834, sensitivity 77.10%, specificity 79.56%) (Figure 2, Table 5).

IL-2R exhibited strong discriminative power (AUC = 0.825) and clinical utility. Its high specificity (85.23%) minimizes false positives, reducing unnecessary aggressive treatment in low-risk patients. IL-2R, composed of α, β, and γ chains, is essential for T-lymphocyte activation and proliferation. Elevated IL-2R levels have been associated with a variety of immune-mediated disorders, as they reflect heightened immune cell activation (25, 26). One of the hallmarks of KD is the immune system’s activation, namely that of T cells (13, 20). Our results show that higher IL-2R levels at admission are significantly associated with IVIG resistance, suggesting that elevated immune cell activation may contribute to IVIG resistance. Our findings are consistent with earlier reports linking abnormal T-cell activation to IVIG resistance and adverse outcomes (14), thereby reinforcing IL-2R as a mechanistically plausible and clinically valuable predictor.

IL-10 was an independent predictor but with modest clinical impact (OR = 1.137), meaning a 1 pg/mL increase in IL-10 is associated with a 13.7% increased risk of IVIG resistance. Consistent with subgroup analysis results, IL-10 serves as a reliable predictor in infants under 12 months, where IL-2R shows limited predictive value. Its higher sensitivity (78.23%) complements IL-2R’s specificity, making it useful for identifying at-risk patients in younger cohorts (<12 months) where IL-2R’s performance is less consistent.

In contrast to the T-cell activation marker IL-2R, we also investigated the anti-inflammatory cytokine IL-10. IL-10 is a key anti-inflammatory cytokine that regulates the immune response by inhibiting the synthesis of pro-inflammatory cytokines (27). It is believed that a major contributing factor to vasculitis in KD is an imbalance of pro-inflammatory and anti-inflammatory cytokines (28). Some investigations have indicated elevated IL-10 levels in KD patients, suggesting its role in the illness process (18, 29). Although the odds ratio for IL-10 was modest (AUC = 0.767), its combination with IL-2R significantly enhanced predictive accuracy. Our research shows that IVIG resistance is linked to greater IL-10 levels upon admission, suggesting a dysregulated anti-inflammatory response in non-responders.

These findings support the hypothesis that IVIG resistance is partially driven by aberrant immune responses (30). IL-10 deficiency has been shown to significantly influence antibody sialylation patterns, particularly through the upregulation of α2, 3-linked sialylation, which may interfere with the anti-inflammatory effects of IVIG therapy (31, 32). Notably, this IL-10-mediated disruption of antibody sialylation may reduce the ability of IVIG to modulate the immune response, thereby contributing to IVIG resistance in KD patients.

Complementing the pro-inflammatory signal captured by IL-2R, the anti-inflammatory cytokine IL-10 also demonstrated independent predictive value, albeit with a different performance profile. Although the absolute increase in AUC was modest, the combined model achieved a more favorable balance between sensitivity and specificity, which might be clinically preferable in some settings where missing a potential non-responder is of greater concern.

Other cytokines, including IL-6, IL-8, and TNF-α, were elevated in resistant patients but did not retain independent predictive significance. This suggests that although these mediators contribute to the inflammatory milieu of KD, their predictive value is less specific than IL-2R and IL-10. Prior studies have reported inconsistent associations for IL-6 and TNF-α, further supporting the need to focus on cytokines with stronger mechanistic relevance and reproducibility across populations.

The identification of IL-2R and IL-10 as independent predictors of IVIG resistance has important translational potential. The cutoff values determined in this study (IL-2R ≥ 21.84 U/mL, IL-10 ≥ 2.695 pg/mL) provide clinically applicable thresholds that could guide early intervention. Patients exceeding these thresholds may benefit from prompt consideration of adjunctive therapies, such as corticosteroids or biologics, before the development of severe coronary complications. Integrating cytokine biomarkers into existing risk scores may also improve predictive accuracy, addressing the limited applicability of current models across diverse populations.

Several limitations should be acknowledged. First, this was a single-center retrospective study, and the relatively small number of IVIG-resistant cases (n=88) may limit generalizability. Second, cytokine levels were measured only once at admission, preventing assessment of their temporal dynamics during disease progression or in response to therapy. Third, although key confounders such as age, sex, and weight were adjusted for, other demographic and clinical variables may still influence biomarker performance. Finally, because the cohort was derived from a single Chinese center with uniform treatment protocols, validation in multi-center, ethnically diverse populations under different clinical practices is essential. Given the variations in demographics and second-line treatment strategies across regions (e.g., North America and Japan), further validation in diverse clinical settings is warranted.

Future research should therefore focus on prospective, large-scale validation of IL-2R and IL-10 as predictive biomarkers in pre-IVIG risk stratification frameworks. Serial cytokine measurements at predefined intervals (e.g., pre-IVIG, 24 h, 48 h post-treatment) may also provide insights into the dynamic immune responses underlying resistance. In addition, translational studies exploring how modulation of IL-10 and IL-2R pathways may enhance IVIG efficacy could open new therapeutic avenues. Such studies could identify critical therapeutic window and offer greater insights into the underlying pathological processes.