Risk Factors and Predictive Models for Intravenous Immunoglobulin Resistance in Children with Recurrent Kawasaki Disease

PURPOSE: To explore the risk factors and develop predictive models for intravenous immunoglobulin (IVIG) resistance in children with recurrent Kawasaki disease (KD). PATIENTS AND METHODS: Patients with recurrent KD were retrospectively reviewed. Clinical and laboratory data at recurrence were collected and compared between patients with and without IVIG resistance. The patients were randomly divided into training and validation cohorts for model development and validation. All variables were subjected to standard Lasso and its variant group Lasso analyses, respectively, to construct predictive models. Model performance was evaluated by receiver operating characteristics (ROC) curves, calibration curves, and Hosmer-Lemeshow tests. RESULTS: A total of 90 children with recurrent KD were included. A total of 16 cases were IVIG resistant. The patients with IVIG resistance had higher age and IVIG resistance probability at the first episode, increased CRP levels, neutrophil count, neutrophil percentage, direct bilirubin level, prothrombin time, and international normalized ratio, and decreased lymphocyte count, lymphocyte percentage, and serum sodium levels. Five variables including age and IVIG resistance at the first episode, lymphocytes count, serum sodium levels, and CRP levels were finally selected by standard Lasso (lLasso model) and four variables including age and IVIG resistance at the first episode, neutrophil percentage, and CRP levels were selected by group Lasso (gLasso). ROC curves suggested lLasso and gLasso models had similar excellent discrimination in both the training cohort (0.895 vs 0.906) and the validation cohort (0.855 vs 0.909). Hosmer-Lemeshow tests suggested the two models exerted a good calibration. Two nomograms were also constructed to facilitate the potential application of the two models. CONCLUSION: Age and IVIG resistance at the first episode and some laboratory variables may be risk factors for IVIG resistance in recurrent KD. Two predictive models for IVIG resistance with excellent performance were established in recurrent KD. External validation should be performed before clinical use.