Early identification of intravenous immunoglobulin (IVIG)-resistant Kawasaki disease (KD) is important for making a suitable therapeutic strategy for children with KD.
This study included a training set and an external validation set. The training set included 635 children (588 IVIG-sensitive and 47 IVIG-resistant KD) hospitalized in Wuhan Children’s Hospital, Hubei, China. Univariate analyses and binary logistic regression equation was incorporated to find the associated variables of the IVIG-resistant KD. A scoring model for predicting IVIG-resistant KD was established according to odds ratio (OR) values and receiver operating characteristic curves. The external validation set consisted of 391 children (358 IVIG-sensitive and 33 IVIG-resistant KD) hospitalized in Peking University First Hospital, Beijing, China. The predictive ability of the model of IVIG-resistant KD were externally validated by the real clinically diagnosed KD cases.
Fifteen variables in the training set were statistically different between IVIG-sensitive and IVIG-resistant KD children, including rash, duration of fever, peripheral blood neutrophil-to-lymphocyte ratio (NLR), prognostic nutritional index (PNI), percentage of monocytes and percentage of eosinophils, and serum alanine aminotransferase, aspartate aminotransferase, total bilirubin (TB), direct bilirubin, glutamyl transpeptidase, prealbumin, sodium ion, potassium ion and high-sensitivity C-reactive protein. According to logistic equation analysis, the final three independent correlates to IVIG-resistant KD were serum TB ≥ 12.8 μmol/L, peripheral blood NLR ≥ 5.0 and peripheral blood PNI ≤ 52.4. According to the OR values, three variables were assigned the points of 2, 2 and 1, respectively. When the score was ≥ 3 points, the sensitivity to predict IVIG-resistant KD was 80.9% and the specificity was 77.6%. In the validation set, the sensitivity, specificity and accuracy of the predictive model of IVIG-resistant KD were 72.7%, 84.9%, and 83.9%, respectively.
A scoring model was constructed to predict IVIG-resistant KD, which would greatly assist pediatricians in the early prediction of IVIG-resistant KD.