This retrospective cohort study utilized clinical data from pediatric patients diagnosed with KD complicated by CAA at Guangzhou Women and Children's Medical Center between January 2015 and December 2023. During this period, a total of 140 patients were diagnosed with KD and CAA. The inclusion criteria were: (1) meeting the established diagnostic criteria for KD; (2) age ≤18 years at diagnosis; (3) echocardiographically confirmed CAA during the acute phase, defined as a maximum coronary artery Z-score ≥2.5 in any segment; and (4) availability of complete serial echocardiographic data from both the acute phase (within 1 month of onset) and at the 1-year follow-up. Exclusion criteria included: (1) significant missing clinical data (n = 0), and (2) lost to follow-up or incomplete 1-year follow-up data (n = 5). Ultimately, 135 patients met all criteria and were included in the final analysis. The study design and statistical analysis were conducted collaboratively by investigators from Guangzhou Women and Children's Medical Center and The First Affiliated Hospital of Guangxi Medical University.

To ensure consistent and objective assessments, a rigorous, multi-step adjudication process was implemented for data collection. Two experienced cardiovascular ultrasound specialists, who were blinded to all patient outcomes and clinical data, independently conducted a retrospective review of all stored echocardiographic images and reports from the study period (January 2015 to December 2023). They recalculated coronary artery Z-scores using the 2017 American Heart Association (AHA) formulas. Aneurysm morphology was classified as fusiform (longitudinal dimension > transverse dimension) or saccular (transverse dimension ≥ longitudinal dimension). Any discrepant cases in morphological classification were adjudicated by a third senior specialist. The final morphology and Z-score for each patient were determined by consensus among all three experts. For patients with aneurysms in multiple locations, the lesion with the largest Z-score was designated as the index aneurysm for analysis. Based on this consensus data, the maximum Z-score (ZM) during the early disease phase was designated as the primary predictor. Baseline demographic characteristics, clinical features, and laboratory parameters were collected from routine medical records.

Index Aneurysm: For patients with multiple aneurysms, the single aneurysm with the highest Z-score during the acute phase was designated as the “index aneurysm” for all subsequent analyses. This approach ensured one data point per patient and focused on the most severe lesion. Predictor Variables: Size: The maximum Z-score of the index aneurysm (ZM), treated as a continuous variable. Morphology: The shape of the index aneurysm, classified as fusiform (longitudinal dimension > transverse dimension) or saccular (transverse dimension ≥ longitudinal dimension). Morphology was assessed by reviewing stored echocardiographic images and reports. Location: Aneurysms located in the left anterior descending artery or the right coronary artery. Outcome Definition: The primary outcome was the persistence of the index aneurysm at the 1-year follow-up (±1 month), defined as the index aneurysm maintaining a Z-score ≥2.5.

IVIG non-response was defined as recurrence of fever (axillary temperature ≥38.0 °C) at least 36 h after the completion of the initial IVIG (2 g/kg) infusion, accompanied by persistent or recrudescent clinical signs of KD. Delayed IVIG treatment was defined as the initiation of IVIG therapy later than 10 days from the onset of fever.

According to medical records, all patients with coronary artery aneurysms received antithrombotic therapy based on aneurysm severity. For patients with small aneurysms (ZM: 2.5–5.0), the initial recorded regimen was aspirin. For children with concomitant G6PD deficiency, the recorded treatment was switched to clopidogrel. For patients with medium aneurysms (ZM: 5.0–10.0), records indicated that most received aspirin monotherapy (n = 15). However, 8 patients (34.8%) with high-risk features (e.g., multiple aneurysms, saccular morphology) received combined therapy with aspirin and clopidogrel, as detailed in Table 5. For the patient with a large aneurysm (ZM ≥ 10.0), the medical record indicated treatment with aspirin, clopidogrel, and warfarin. The primary outcome of this study was anatomical persistence of the aneurysm (Z-score ≥2.5). Additionally, the occurrence of Major Adverse Cardiovascular Events (MACE), defined as a composite of myocardial infarction, coronary intervention, or cardiac death, as well as any coronary thrombus formation, was retrospectively reviewed during the one-year follow-up period.

Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed data are presented as mean ± standard deviation and compared using independent samples t-tests, while non-normally distributed data are expressed as median (interquartile range) and compared using Mann–Whitney U tests. Categorical variables are summarized as counts (percentages) and compared using chi-square or Fisher's exact tests as appropriate. Univariate and multivariate logistic regression models were employed to assess independent effects. The predictive performance of the maximum Z-score was evaluated using receiver operating characteristic (ROC) curve analysis. The overall performance of the final CAC model was further assessed by evaluating calibration and clinical utility. A univariate logistic regression analysis was performed on a comprehensive set of baseline characteristics to screen for associations with the primary outcome; the full results are presented in Supplementary Table S1. It should be noted that all reported performance metrics for the CAC model represent “apparent” performance estimates derived from the development dataset, which are likely optimistic and represent a best-case scenario; internal validation using bootstrapping was not performed due to resource constraints and sample size considerations; thus external validation is required to confirm generalizable performance. All analyses were performed using SPSS 26.0 and R 4.2.0, with a two-sided p-value <0.05 considered statistically significant.

Given the limited number of outcome events and to prevent overfitting of the multivariable model, we pre-specified a parsimonious set of predictors based on the study's primary aim. The core predictors were the intrinsic aneurysm characteristics: maximum Z-score (ZM), morphology, and location. Although univariate analysis identified several systemic and clinical factors associated with the outcome (Supplementary Table S1), these were not included in the primary multivariable model. This approach ensures model stability and focuses the analysis on evaluating the incremental value of aneurysm anatomy beyond size alone. The role of systemic factors is acknowledged and discussed in the context of the literature.

In accordance with the Declaration of Helsinki, institutional review board (IRB) approval was sought. The IRB granted exemption from ethical review for this study. As data were anonymized, the requirement for informed consent was waived.

A total of 135 children with Kawasaki disease and coronary artery aneurysms (CAA) were included in the final analysis. Based on the CAA status at the 1-year follow-up, patients were stratified into two groups: the CAA regression group (n = 105) and the CAA persistence group (n = 30). The CAA persistence group demonstrated a significantly higher median ZM during the early phase compared to the regression group (5.69 vs. 3.19, P < 0.001). A comprehensive comparison of all baseline clinical and demographic characteristics between the two groups is detailed in Table 1. Significant intergroup differences (P < 0.05) were also observed in multiple baseline parameters, including age, height, total fever duration, laboratory findings (white blood cell count, platelet count, absolute neutrophil count, NT-proBNP, chloride, total protein, globulin, albumin-to-globulin ratio), and the proportion of patients who received delayed intravenous immunoglobulin treatment. These baseline differences suggest that children in the CAA persistence group may have experienced a more severe inflammatory response and overall disease severity during the acute phase, which is consistent with previous research. As this study primarily aims to investigate the predictive value of the aneurysms' own imaging characteristics, the subsequent analysis focused on the three core factors: ZM, morphology, and location. The results of the univariate analysis for all baseline characteristics are provided in Supplementary Table S1.

Univariable logistic regression identified the ZM within the first month as a powerful predictor of persistent coronary aneurysms at 1 year (OR = 4.925, 95% CI: 2.738–8.856, P < 0.001). Receiver operating characteristic (ROC) analysis demonstrated excellent discriminatory power with an area under the curve (AUC) of 0.909 (95% CI: 0.848–0.970; Figure 1).

The baseline distribution of the key aneurysm characteristics is summarized in Table 2. Saccular morphology was present in 20.7% of the cohort, while LAD involvement was observed in 41.5% of patients. Both characteristics demonstrated a strong bivariate association with the outcome, being markedly more prevalent in the persistence group (46.7% and 60.0%, respectively) than in the regression group (13.3% and 36.2%, P < 0.001 for morphology, P = 0.034 for location). Furthermore, a larger initial aneurysm size, as measured by the ZM, was a prominent feature in the persistence group (Median: 5.69, IQR: 4.85–6.29) compared to the regression group (Median: 3.19, IQR: 2.69–4.11, P < 0.001).

The results of the univariable and multivariable logistic regression analyses are presented in Table 3. In univariable analysis, larger ZM, saccular morphology, and LAD involvement were all significantly associated with aneurysm persistence (all P < 0.05). In the multivariable model that included all three characteristics, larger ZM (adjusted OR = 6.775, 95% CI: 3.133–14.648, P < 0.001) and LAD involvement (adjusted OR = 4.304, 95% CI: 1.163–15.928, P < 0.05) emerged as independent predictors with precise effect estimates.

Saccular morphology also demonstrated a significant association after adjustment (adjusted OR = 7.648, 95% CI: 1.428–40.967, P < 0.05). However, the considerable widening of its confidence interval indicates instability in the point estimate, likely reflecting limited statistical power within this subgroup. Therefore, while these results support its role as an independent risk factor, the exact magnitude of its effect remains uncertain. Given this statistical imprecision alongside its strong unadjusted association, we further explored its clinical relevance in the high-risk subgroup analysis below.

The results of the model performance assessments are presented in Figure 2. The discriminatory ability of the four models was evaluated using ROC analysis (Figure 2A). The model based solely on the maximum Z-score (ModA) served as the baseline, achieving an AUC of 0.909 (95% CI: 0.848–0.970). The multivariable models, which incorporated additional aneurysm characteristics, demonstrated progressively higher AUC values: the model combining ZM and morphology (ModB) achieved an AUC of 0.924 (95% CI: 0.872–0.976); the model combining ZM and location (ModC) achieved an AUC of 0.928 (95% CI: 0.867–0.990); and the comprehensive model incorporating ZM, morphology, and location (ModD) achieved the highest AUC of 0.941 (95% CI: 0.896–0.986). DeLong's test for correlated ROC curves indicated that the superiority of the comprehensive model (ModD) over the baseline ModA was statistically significant (difference in AUC = 0.032, 95% CI for difference: 0.004–0.060, P = 0.025). In contrast, the improvements of ModB (difference in AUC = 0.015, P = 0.347) and ModC (difference in AUC = 0.019, P = 0.118) over ModA were not statistically significant. The calibration curve (Figure 2B) showed that the predicted probabilities of the primary model were in close agreement with the observed outcomes across the entire range of risk. Decision curve analysis (Figure 2C) was employed to evaluate clinical utility. All four models provided a greater net benefit than the “treat all” or “treat none” strategies across a wide range of clinically reasonable threshold probabilities. Among them, the comprehensive model (ModD) yielded the highest net benefit, solidifying its value for informing clinical decisions.

The longitudinal evolution of ZM from the acute phase to 1-year follow-up is summarized in Figure 3. Patients with CAA regression demonstrated a pronounced downward trend in median Z-scores over time, with values decreasing from 3.19 (IQR: 2.65–4.23) at baseline to 1.42 (IQR: 1.23–1.54) at 12 months. In contrast, the CAA persistence group maintained elevated Z-scores throughout the follow-up period, with median values of 5.69 (IQR: 4.86–6.88) at baseline and 4.80 (IQR: 3.69–5.42) at 12 months.

We conducted a descriptive analysis on the high-risk subgroup of patients with a maximum Z-score (Zmax) ≥ 5.0 (n = 24). Owing to the limited sample size of this subgroup (n = 24), conducting a stable multivariable analysis was not feasible. The following findings should therefore be interpreted as exploratory and hypothesis-generating. The baseline characteristics and aneurysm features of this subgroup, stratified by persistence outcome at one year, are detailed in Table 4. Within this high-risk cohort, 10 patients (41.7%) had persistent aneurysms. The persistence group had a significantly larger initial aneurysm size (Median Zmax: 7.46 vs. 5.76, P < 0.001). Consistent with the signal of elevated risk observed in the primary multivariable model, saccular morphology was overwhelmingly more prevalent in the persistence group (90.0%) compared to the regression group (14.3%, P = 0.001). Involvement of the left anterior descending artery (LAD) was also more common in the persistence group (70.0% vs. 35.7%), although this difference did not reach statistical significance (P = 0.214). No other baseline clinical characteristics showed significant differences between the groups.

The antithrombotic regimens for the entire cohort are summarized in Table 5. In brief, among the 111 patients with small aneurysms, 12 with G6PD deficiency received clopidogrel instead of aspirin. Of the 23 patients with medium aneurysms, 8 received dual antiplatelet therapy with aspirin and clopidogrel. The single patient with a giant aneurysm was on a triple therapy regimen of aspirin, clopidogrel, and warfarin.

No Major Adverse Cardiovascular Events (MACE) were documented during the follow-up period. However, coronary thrombus formation was identified in 2 patients. The detailed clinical characteristics and management of these thrombosis cases are presented in Table 6. Both patients were treated with a protocol of urokinase thrombolysis followed by a heparin sodium bridge, culminating in long-term warfarin therapy. Medical records indicated that the thrombi were controlled in both instances following these interventions.