Establishment and validation of clinical prediction model for steroid-resistant nephrotic syndrome in children

Abstract

Objective: To identify the clinically relevant factors of steroid-resistant nephrotic syndrome (SSNS) in children and establish a predictive model followed by verifying its feasibility. Methods: A retrospective analysis was performed in a total of 111 children with nephrotic syndrome admitted to Children's Hospital of ShanXi from January 2016 to December 2021. The clinical data of general conditions, manifestations, laboratory tests, treatment, and prognosis were collected. According to the steroid response, patients were divided into SSNS and steroid resistant nephrotic syndrome (SRNS) group. Single factor Logistic regression analysis was used for comparison between the 2 groups, and variables with statistically significant differences were included in multivariate Logistic regression analysis. The multivariate Logistic regression analysis was used to identify the related variables of children with SRNS. The area under the receiver operating characteristic curve (ROC), the calibration curve and the clinical decision curve were used to evaluate its effectiveness of the variables. Results: Totally 111 children with nephrotic syndrome was composed of 66 boys and 45 girls, aged 3.2 (2.0, 6.6) years. There were 65 patients in the SSNS group and 46 in the SRNS group.Univariate Logistic regression analysis showed that the 6 variables, including erythrocyte sedimentation rate, 25-hydroxyvitamin D, suppressor T cells, D-dimer, fibrin degradation products, β2-microglobulin, had statistically significant differences between SSNS and SRNS groups (85 (52, 104) vs. 105 (85, 120) mm/1 h, 18 (12, 39) vs. 16 (12, 25) nmol/L, 0.23 (0.19, 0.27) vs. 0.25 (0.20, 0.31), 0.7 (0.6, 1.1) vs. 1.1 (0.9, 1.7) g/L, 3.1 (2.3, 4.1) vs. 3.3 (2.7, 5.8) g/L, 2.3 (1.9,2.8) vs. 3.0 (2.5, 3.7) g/L, χ2=3.73, -2.42, 2.24, 3.38, 2.24,3.93,all P<0.05), were included in the multivariate Logistic regression analysis. Finally, we found that 4 variables including erythrocyte sedimentation rate, suppressor T cells, D-dimer and β2-microglobulin (OR=1.02, 1.12, 25.61, 3.38, 95%CI 1.00-1.04, 1.03-1.22, 1.92-341.04, 1.65-6.94, all P<0.05) had significant correlation with SRNS. The optimal prediction model was selected. The ROC curve cut-off=0.38, with the sensitivity of 0.83, the specificity of 0.77 and area under curve of 0.87. The calibration curve showed that the predicted probability of SRNS group occurrence was in good agreement with the actual occurrence probability, χ2=9.12, P=0.426. The clinical decision curve showed good clinical applicability. The net benefit is up to 0.2. Make the nomogram. Conclusions: The prediction model based on the 4 identified risk factors including erythrocyte sedimentation rate, suppressor T cells, D-dimer and β2-microglobulin was suitable for the early diagnosis and prediction of SRNS in children. The prediction effect was promising in clinical application.