Implications of Changing Z-Score Models for Coronary Artery Dimensions in Kawasaki Disease

Background: Coronary artery abnormalities in Kawasaki disease (KD) are assessed using echocardiographic z-scores. We hypothesized that changing the coronary artery (CA) z-score model would alter diagnosis and management of children with KD.

Methods: In this retrospective single-center study of children treated for KD (9/2007-1/2020), we collected echocardiographic measurements for the left anterior descending (LAD), right (RCA), and left main (LMCA) coronary arteries during 3 illness phases and calculated Boston and Pediatric Heart Network (PHN) z-scores. Agreement between Boston and PHN z-scores was assessed using Kappa (κ) and Lin's Concordance Correlation Coefficients (CCC) and Bland-Altman analysis.

Results: For 904 echocardiograms from 357 children, the median Boston LAD z-score was lower than the PHN (0.3 [IQR - 0.6, 1.5] vs 1.6 [IQR 0.7, 2.8], CCC 0.94 [95% CI 0.93, 0.95], moderate agreement), aggregated across all illness phases. RCA and LMCA z-scores showed substantial agreement. With conversion from Boston to PHN models, the percentage of individual LAD z-scores ≥ 2.5 increased (14.6% to 32.1%). At least one CA z-score classification changed in 213 children (59.7%) across all phases, and 48 children (13.4%) had a change that altered recommended antithrombotic strategy. Agreement between models differed by age, sex, and race.

Conclusions: Conversion from Boston to PHN z-scores changed at least 1 CA z-score classification in over half of KD patients and changed recommended antithrombotic management in 13%, largely driven by LAD measurements. Since diagnosis and management of KD and KD-like diseases rely upon CA z-scores, the clinical and research implications of these findings merit further exploration.