The model was the first to incorporate C-reactive protein/albumin ratio and apply LASSO regression analysis to a predictive model for children with Kawasaki disease.
A nomogram model can accurately predict intravenous immunoglobulin (IVIG) resistance in patients with Kawasaki disease (KD), and C-reactive protein/albumin ratio (CAR) was an important factor in predicting this resistance, according to the results of a study published in Frontiers in Cardiovascular Medicine.
The first-line treatment of KD is high-dose IVIG combined with aspirin, which reduces the incidence of coronary artery lesions (CALs) from 20%-25% to 2%-4%. However, initial treatment with IVIG is not effective in 7.5%-26.8% of children with KD who remain at risk for developing CALs.
Prior studies have established region-specific predictive scoring models based on the clinical data of children with KD to conduct risk assessments. Despite this, the small sample size in those studies, the choice of modeling methods, and the lack of internal and external validations have led to those models not meeting clinical expectations.
The aim of the current study was to collect and examine clinical data and combining indicators of KD to develop a new predictive scoring model for IVIG-resistant KD to facilitate early assessment and treatment. Variables that were featured in the model included demographic characteristics, clinical outcomes, laboratory test results such as CALs, C-reactive proteins (CRPs), and white blood cell (WBC) count.
Data for 1259 patients were included into the training set, and data for 539 patients were included in the validation set for internal validation. Of these, 113 (6.3%) cases were defined as IVIG non-responders. Nine predictors were selected for the training set, including CAR, CRP, and sodium.
LASSO regression analysis was used to screen the previously mentioned predictor variables, and using multivariate logistic regression, a prediction model was established. Variables that had nonzero coefficients of the LASSO regression model were included: lymphocyte percentage (LY%), CAR, aspartate aminotransferase (AST), sodium, and total bilirubin (TB).
The researchers created ModA and ModB receiver operating characteristic (ROC) curves and compared them. The net reclassification index (NRI) of ModA was significantly better than that of ModB (0.304; 95% CI, 0.081-0.527, p < 0.05), so the investigators chose the ModA to predict IVIG resistance in KD.
The area under the ROC (AUC) of the training set was 0.825 (95% CI, 0.781-0.869), and the sensitivity and specificity of were 0.723 and 0.744, respectively. The AUC of the internal validation set was 0.791 (95% CI, 0.694-0.890), and the AUC of the prospective external validation was 0.801 (95% CI, 0.717-0.885).
This research method employed by the investigators was the first to apply LASSO regression to select variables in the data set, which reduced the dimension of the data and eliminated the collinearity between variables, and then used logistic regression to determine immunoglobulin resistance.
Furthermore, the investigators wrote that their study is the first to include CAR in the prediction model. Prior studies have found that this ratio is associated with the formation of CAL, which is related to inflammatory processes in children with KD.
The study has several limitations that were noted by the researchers. Firstly, the study uses a retrospective design to establish the model, and as a result, the investigators could not explore and determine some relevant factors, such as components of IVIG and potential biomarkers.
Additionally, the results of echocardiography were assessed by different cardiologists, which could have introduced human error. The investigators wrote that a more detailed investigation will be performed in a multicenter trial to further analyze factors that affect IVIG components in a larger sample of children with KD.
“Nevertheless, based on this large cohort study, reasonable statistical methods, and the results of internal and external validation, we believe that our model can predict IVIG resistance in KD patients in our region,” the study authors concluded.
Wang S, Ding C, Zhang Q, et al. A novel model for predicting intravenous immunoglobulin-resistance in Kawasaki disease: a large cohort study. Front Cardiovascular Med. 2023;10. doi:10.3389/fcvm.2023.1226592