Research On Predicting Model Of Intravenous Immunoglobulin Resistance In Kawasaki Disease

Objective To establish a higher sensitive model predicting IVIG unresponsive in KD by combing laboratory indicators and gene.Methods 330 KD patients and 105 healthy children treated in Shanghai Children′s Hospital between November 2015 and August 2018 were selected.Clinical indicators were found differently between IVIG unresponsive and IVIG responsive group using electronic database.A next generation sequencing technology was performed to screen the gene associated with IVIG resistance in KD.Subsequently,a prediction model of IVIG resistance which combined the w GRS of SNP loci with the laboratory data was built using a random forest classifier.Meanwhile,Kobayashi score was used to score the subjects recruited in the study and compared the sensitivity and specificity with the new predicting model.Results 10 clinical indicators with P value less than 0.05 and lacking data less than 30% were found.Subjects who were IVIG unresponsive,compared with subjects who were responsive to IVIG therapy,were similar in age and sex,in univariate analysis of baseline clinical indicators,more likely to have higher C-reactive protein level and percentage of neutrophils.Subjects who were IVIG unresponsive also had lower platelet count,erythrocyte sedimentation rate,serum albumin concentration and hemoglobin level compared with subjects who were responsive to IVIG.There were 25 SNPs that showed significantly difference between IVIG unresponsive and IVIG responsive group(P<0.05).The sensitivity and specificity of the new model which combined clinical indicators and gene were 80% and 87.1%.However,the sensitivity of the Kobayashi score to predict IVIG resistance in this study was only 34.4%.Conclusions Gene can enhance the sensitivity of the scoring system to predict IVIG unresponsive in KD,and it is valuable to assist the clinical diagnose and treatments.