Risk Model Development and Validation for Prediction of Coronary Artery Aneurysms in Kawasaki Disease in a North American Population

Abstract

Background Accurate prediction of coronary artery aneurysms ( CAAs ) in patients with Kawasaki disease remains challenging in North American cohorts. We sought to develop and validate a risk model for CAA prediction. Methods and Results A binary outcome of CAA was defined as left anterior descending or right coronary artery Z score ≥2.5 at 2 to 8 weeks after fever onset in a development cohort (n=903) and a validation cohort (n=185) of patients with Kawasaki disease. Associations of baseline clinical, laboratory, and echocardiographic variables with later CAA were assessed in the development cohort using logistic regression. Discrimination (c statistic) and calibration (Hosmer-Lemeshow) of the final model were evaluated. A practical risk score assigning points to each variable in the final model was created based on model coefficients from the development cohort. Predictors of CAAs at 2 to 8 weeks were baseline Z score of left anterior descending or right coronary artery ≥2.0, age <6 months, Asian race, and C-reactive protein ≥13 mg/ dL (c=0.82 in the development cohort, c=0.93 in the validation cohort). The CAA risk score assigned 2 points for baseline Z score of left anterior descending or right coronary artery ≥2.0 and 1 point for each of the other variables, with creation of low- (0-1), moderate- (2), and high- (3-5) risk groups. The odds of CAA s were 16-fold greater in the high- versus the low-risk groups in the development cohort (odds ratio, 16.4; 95% CI , 9.71-27.7 [ P<0.001]), and >40-fold greater in the validation cohort (odds ratio, 44.0; 95% CI, 10.8-180 [ P<0.001]). Conclusions Our risk model for CAA in Kawasaki disease consisting of baseline demographic, laboratory, and echocardiographic variables had excellent predictive utility and should undergo prospective testing.