Predicting Coronary Artery Aneurysms in Kawasaki Disease at a North American Center: An Assessment of Baseline z Scores

Background-Accurate risk prediction of coronary artery aneurysms (CAAs) in North American children with Kawasaki disease remains a clinical challenge. We sought to determine the predictive utility of baseline coronary dimensions adjusted for body surface area (z scores) for future CAAs in Kawasaki disease and explored the extent to which addition of established Japanese risk scores to baseline coronary artery z scores improved discrimination for CAA development. Methods and Results-We explored the relationships of CAA with baseline z scores; with Kobayashi, Sano, Egami, and Harada risk scores; and with the combination of baseline z scores and risk scores. We defined CAA as a maximum z score (zMax) >= 2.5 of the left anterior descending or right coronary artery at 4 to 8 weeks of illness. Of 261 patients, 77 patients (29%) had a baseline zMax >= 2.0. CAAs occurred in 15 patients (6%). CAAs were strongly associated with baseline zMax >= 2.0 versus <2.0 (12 [16%] versus 3 [2%], respectively, P<0.001). Baseline zMax >= 2.0 had a C statistic of 0.77, good sensitivity (80%), and excellent negative predictive value (98%). None of the risk scores alone had adequate discrimination. When high-risk status per the Japanese risk scores was added to models containing baseline zMax >= 2.0, none were significantly better than baseline zMax >= 2.0 alone. Conclusions-In a North American center, baseline zMax >= 2.0 in children with Kawasaki disease demonstrated high predictive utility for later development of CAA. Future studies should validate the utility of our findings.