Most studies of coronary artery involvement and associated risk factors in Kawasaki disease have used the Japanese Ministry of Health dichotomous criteria. Analysis of serial normalized artery measurements may reveal a broader continuous spectrum of involvement and different risk factors. Clinical, laboratory, and echocardiographic measurements obtained at baseline and 1 week and 5 weeks after presentation were examined in 190 Kawasaki disease patients as part of a clinical trial of primary therapy with pulse steroids in addition to standard intravenous immunoglobulin. Maximum coronary artery z score normalized to body surface area was significantly greater than normal at all time points, decreasing significantly over time from baseline. A maximal z score > or = 2.5 at any time was noted in 26% of patients. Japanese Ministry of Health dimensional criteria were met by 23% of patients. Significant independent factors associated with greater z score at any time included younger patient age, longer interval from disease onset to treatment with intravenous immunoglobulin, lower serum IgM level at baseline, and lower minimum serum albumin level. z scores of the proximal right coronary artery were higher than those in the left anterior descending branch. Analyses of serial normalized coronary artery measurements in optimally treated Kawasaki disease patients demonstrated that for most patients, measurements are greatest at baseline and subsequently diminish; baseline measurements appear to be good predictors of involvement during early follow-up. When a more precise assessment is used, risk factors for coronary artery involvement are similar to those defined with arbitrary dichotomous criteria.