# Value of dynamic changes in inflammatory biomarkers for predicting intravenous immunoglobulin resistance in children with Kawasaki disease

## TL;DR

This study shows that changes in blood markers before and after treatment can help predict which children with Kawasaki disease will not respond to IVIG therapy.

## Contribution

The study introduces fractional changes in inflammatory biomarkers as novel predictors of IVIG resistance in Kawasaki disease.

## Key findings

- WBC and NE count fractional changes were the strongest predictors of IVIG resistance with AUCs of 0.7677 and 0.7818.
- Combined AUC of FC reached 0.8307 in the training cohort and 0.8564 in the validation cohort.
- Nonlinear relationships between predictors and IVIG resistance were confirmed via RCS analysis.

## Abstract

This study assessed the predictive value of dynamic laboratory parameter changes before and after intravenous immunoglobulin (IVIG) treatment for IVIG resistance in children with Kawasaki disease (KD).

Children with KD were stratified based on the occurrence of IVIG resistance. Logistic regression analyses were conducted to identify independent risk factors. The predictive performance of variables and their fractional changes (FC) was evaluated through receiver operating characteristic (ROC) curve analysis. Nonlinear associations between predictors and outcomes were examined via restricted cubic spline (RCS) analysis.

The Soochow cohort analyzed 1,796 children, with IVIG resistance observed in 140 cases (7.8%). 636 children from the Anhui cohort were included in external validation. Multivariate regression analysis identified pre-treatment CLR and Hb, post-treatment CLR, LMR, NLR, Hb, and FCs in WBC, Hb, NE%, and NE count as significant independent predictors of IVIG resistance (P < 0.05). ROC analysis demonstrated that WBC(FC) and NE count(FC) were the strongest predictors of IVIG resistance, with AUCs of 0.7677 and 0.7818, respectively, outperforming other parameters. The combined AUC of FC was 0.8307 in the Soochow cohort and 0.8564 in the validation cohort. RCS analysis revealed significant nonlinear relationships between predictors and IVIG resistance.

Fractional changes in WBC and NE count were established as robust predictors of IVIG resistance in KD. Future efforts should focus on developing predictive models with thresholds and dynamic risk assessments at various time points to enhance the accuracy of IVIG resistance prediction. Clinicians should closely monitor children with IVIG resistance risk factors and reassess the risk after first treatment.