Predictive value of serum ferritin and a new predictive model for intravenous immunoglobulin resistance in Kawasaki disease

Abstract

Objective: To evaluate the predictive value of serum ferritin (SF) and construct a novel predictive model for intravenous immunoglobulin (IVIG)-resistant Kawasaki disease (KD) (IVIGRKD). Methods: The clinical data of 422 children with KD from January 2017 to December 2019 in Quanzhou Women's and Children's Hospital were retrospectively analyzed. According to the response to IVIG, they were divided into IVIG-resistant group and IVIG-sensitive group. Forty-one clinical indicators including general characteristics and laboratory results were compared between the two groups. Comparisons between groups were performed with unpaired Student t test or Mann-Whitney U test or chi-square test. Receiver operating characteristic (ROC) curve was applied to evaluate the predictive value of SF for IVIGRKD. Binary Logistic regression analysis was used to test whether SF was an independent risk factor for IVIGRKD. Meanwhile, a novel predictive scoring system was established. The comparisons between the new predictive scoring system with four commonly used prediction scoring systems were conducted. Results: A total of 422 KD cases (285 males and 137 females, 17.0 (9.0,29.0) years of age) were enrolled and divided into IVIG-resistant group (n=57) and IVIG-sensitive group (n=365). Seventeen clinical indicators differed significantly between the two groups. SF level of the IVIG-resistant group was significantly higher than that of the sensitive group (245.0 (131.0, 519.0) vs. 145.0 (92.5, 232.5) μg/L, Z=-5.109, P<0.05). ROC curve showed that the Youden index of SF for predicting IVIGRKD was 0.326 (cutoff value 403.5 μg/L). Binary logistic regression analysis showed that SF, days of illness at initial IVIG treatment, cervical lymphadenopathy, pleomorphic rash, white blood cell, C-reactive protein (CRP), activated partial thromboplastin time (APTT), alanine transaminase (ALT) and creatinine were independent risk factors for IVIGRKD. A novel prediction model was constructed, and the cutoff points and score points were as follows: pleomorphic rash, 2 points; cervical lymphadenopathy, 1 point; SF≥ 403.5 μg/L, 1 point; white blood cell ≥ 18.3×109/L, 1 point; CRP≥83.1 mg/L, 1 point; APTT≥25.3 s, 1 point; ALT≥37.5 U/L, 1 point. And patients with scores of 4 or more were at high-risk for IVIGRKD. The Youden index of the four commonly used scoring systems ranged from 0.315 to 0.512. However, the Youden index of the new scoring system was 0.695 (sensitivity 0.772, specificity 0.923) and was the highest among the five scoring systems. Conclusions: SF shows well predictive efficiency for IVIGRKD and is an independent risk factor for IVIGRKD. SF can be used as a new predictor of IVIGRKD.