# Predictive value of aneurysm characteristics for one-year persistence in Kawasaki disease: a retrospective cohort study

## TL;DR

This study shows that the size, shape, and location of coronary aneurysms in children with Kawasaki disease can predict whether the aneurysms will persist after one year.

## Contribution

The study introduces a new model combining aneurysm size, shape, and location to improve prediction of aneurysm persistence in Kawasaki disease.

## Key findings

- The maximum Z-score (ZM) was a strong predictor of persistent aneurysms at one year (AUC: 0.909).
- The CAC model, which includes ZM, morphology, and location, improved prediction accuracy (AUC: 0.941).
- Saccular morphology was more common in persistent aneurysms among high-risk patients (ZM ≥ 5.0).

## Abstract

While systemic factors influence Kawasaki disease outcomes, this study specifically determines the independent and incremental prognostic value of the coronary aneurysm's own characteristics—maximum Z-score (ZM), morphology, and Location—for predicting persistence one year after onset.

This retrospective cohort study enrolled 135 children with KD and coronary artery aneurysms (CAA). We analyzed the maximum Z-score (ZM), morphology (fusiform/saccular), and coronary artery Location (Left/Right/Bilateral) of the index aneurysm (the largest by Z-score). Univariable and multivariable logistic regression were used to identify independent predictors. The predictive performance of a model containing only ZM was compared to that of the Comprehensive Aneurysm Characteristic (CAC) model, which incorporates ZM, morphology, and Location, by assessing the area under the receiver operating characteristic curve (AUC). A descriptive analysis was additionally performed on a high-risk subgroup defined by a ZM ≥ 5.0.

The early ZM was a strong predictor of persistent coronary aneurysms at one year (OR = 4.925, P < 0.001), with an AUC of 0.909. In the multivariable analysis, larger ZM (aOR = 6.775, 95% CI: 3.133–14.648, P < 0.001), saccular morphology (aOR = 7.648, 95% CI: 1.428–40.967, P < 0.05), and LAD involvement (aOR = 4.304, 95% CI: 1.163–15.928, P < 0.05) emerged as independent predictors. The CAC model demonstrated a statistically significant improvement in predictive ability over the ZM-only model (AUC: 0.941 vs. 0.909, P = 0.025). Exploratory analysis of the high-risk subgroup (ZM ≥ 5.0) revealed a markedly higher prevalence of saccular morphology in patients with persistent aneurysms, suggesting it may serve as a crucial risk signal independent of absolute size in this population. The CAC model also showed excellent calibration and superior clinical utility across a wide range of decision thresholds.

The intrinsic characteristics of a coronary aneurysm—its size, shape, and distribution—collectively provide powerful, independent prediction of its persistence at one year. The strong association of specific morphological features with persistence in the high-risk subgroup underscores the value of anatomy-based assessment for refining risk stratification, complementing evaluations based on systemic host factors.