Background Our goals were to find and grade children with multisystem inflammatory syndrome (MIS‐C) by clinical severity and the degree of different system involvement based on multivariate clinical laboratory data and identify the best predictors and scores for disease prognosis and risk stratification. Methods We enrolled 51 patients with confirmed MIS‐C from a single center, and 34 general laboratory parameters and markers were included. Results Using the K ‐means clustering method on kernel principal component analysis (K‐PCA) projections, we identified that our MIS‐C patients could be separated into three clusters. Children belonging to Cluster 3 have the highest levels of ferritin, LDH, ASAT, ALAT, GGT, total bilirubin, direct bilirubin, azotemia, urea, and creatinine, followed by Cluster 2. In contrast, Cluster 1 showed the lowest levels, p < 0.05. Children belonging to Clusters 3 and 2 also needed inotropic support significantly more frequently than Cluster 1 (30% and 10% vs 0%, Fisher exact test p = 0.04). Furthermore, respiratory distress was found only in patients of Clusters 2 and 3 ( p = 0.002). Regarding liver involvement, Clusters 2 and 3 more frequently had cholestasis (61% and 75% vs. 28%, p = 0.012), whereas Cluster 3 was more prominently characterized by an enlarged liver (44% vs. 0% and 5%, p = 0.004). Therefore, our clusters represent different grades, and liver and kidney involvement stages are graded by severity. From ROC curve analysis and several logistic regressions, we identified that age equal to or higher than 9 years old and ferritin levels higher than 470 ng/mL could help distinguish children with more severe kidney and liver involvement with a high probability (ROC AUC = 77 % , p = 0.03 and ROC AUC = 75 % , p = 0.04) approximately equal to the discriminatory potential of creatinine, urea, GGT and total bilirubin. Moreover, levels of LDH with a cutoff of 363 U/l could identify the children within Cluster 3 (ROC AUC = 80 % , p = 0.004). Conclusions LDH and ferritin, as well as age into consideration, could help the clinical assessment of the underlying severity.
Vasilev et al. (Thu,) studied this question.