Comparison of risk scores for predicting intravenous immunoglobulin resistance in Taiwanese patients with Kawasaki disease

Background/Purpose: Kawasaki disease (KD) is the most common type of acquired heart disease in children, and intravenous immunoglobulin (IVIG) therapy is the preferred treatment. Several risk scoring systems have been developed to predict IVIG resistance, which is important in KD management, including the Kobayashi, Egami, and Formosa scores. We evaluated the performance of these scoring systems with a KD patient cohort from Taiwan. Methods: We retrospectively analyzed the medical records of all KD patients admitted to our institution from 2012 to 2017. We compared the characteristics of IVIG-resistant and nonresistant patients and evaluated the predictive ability of the scoring systems for IVIG resistance. Results: We included 84 patients, with 73 receiving IVIG therapy. Eight patients were unresponsive to the first IVIG course. Compared to those with good response to therapy or spontaneous improvement, IVIG-resistant patients had a higher C-reactive protein level (16.1 mg/dL vs. 8.6 mg/dL, p < 0.001), higher percentage of segmented leukocytes (75.7% vs. 61.7%, p = 0.008), and lower albumin level (2.98 mg/dL vs. 3.78 mg/dL, p = 0.001). In determining IVIG resistance, the sensitivity and specificity varied among scoring systems (Kobayashi, 37.5% and 86.8%; Egami, 37.5% and 84.2%; and Formosa, 87.5% and 73.7%, respectively). The positive and negative predictive values of the Formosa score were 25.9% and 98.2%, respectively. Conclusion: The Formosa score had the highest sensitivity in determining IVIG resistance. Although the positive predictive value was low, the negative predictive value could reach 98.2%. The Formosa score was superior to other scoring systems in predicting IVIG resistance in Taiwanese KD patients. Copyright 2020, Formosan Medical Association. Published by Elsevier Taiwan LLC. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).