Anatomical distribution pattern of coronary artery lesions in Kawasaki disease and the prognostic value of dynamic Z-scores for recovery time: a retrospective cohort study

Objective To explore the anatomical distribution of coronary artery lesions (CALs) in Kawasaki disease (KD) and assess the predictive value of Z-scores at different time windows for coronary recovery time. Methods This single-center retrospective cohort included 569 KD children diagnosed in 2021, of whom 211 had coronary dilation. Echocardiography measured Z-scores of five coronary segments. Follow-up was scheduled at diagnosis, week 2, month 1, month 3, 6 months, 1 year, and beyond. Generalized estimating equations tested inter- coronary segment correlations, and Cox regression evaluated the effect of the cumulative maximum Z-score at each time window on recovery time. Model fit was compared using AIC. Results Lesions exhibited a statistical pattern of adjacent-segment correlation along the same vessel's longitudinal axis, strongest within the LAD and RCA systems. AIC comparison revealed that Zmax-T2 (within 2 weeks after diagnosis) provided clinically useful early prognostic information, with hazard ratios numerically similar to those of later time windows (ranging from 0.50 to 0.52). For Zmax-T2, each 1-unit increase corresponded to an approximately 49.6% lower hazard of coronary normalization (HR≈0.504). Compared with single- coronary segment involvement, recovery was significantly slower beginning in the two- coronary segment category. Female sex was associated with faster recovery (HR = 1.369, P = 0.041), whereas age and season showed no significant effect. Conclusion KD coronary lesions exhibit axial adjacent-segment correlation. Zmax-T2 is a practically useful early predictor of recovery, supporting early risk stratification. The number of involved coronary segments remained associated with recovery after adjustment for Zmax-T2, although the proportional hazards assumption was not fully met for this variable, and the association should be interpreted as an average effect over the entire follow-up period.