Cross-Population Validation of the Pediatric CKD Risk-Prediction Tool

Abstract

Introduction: This study aimed to externally validate the performance of the kidney replacement therapy (KRT) risk prediction calculator for chronic kidney disease (CKD) in children in an ethnically distinct East Asian population. Methods: We externally validated the KRT risk prediction calculator for CKD in children using data from the KoreaN Cohort Study on Outcomes in Pediatric CKD (KNOW-Ped CKD) cohort. Six parametric survival models from the generalized gamma family were tested, stratified into 2 groups as follows: change in glomerular filtration rate (GFR)-based (group 1) and cross-sectional (group 2). Missing data (# 7.9%) were addressed via multiple imputations using chained equations. Outcomes were timed to KRT initiation. The model performance was evaluated based on goodness-of-fit, discrimination ability, calibration, and predictive ability. Results: Overall, 533 children were included in the validation cohort. The median age and baseline estimated GFR (eGFR) were 10.8 years (interquartile range [IQR]: 5.3-14.5) and 57.6 ml/min per 1.73 m 2 (IQR: 34.8-81.4), respectively. Over a median follow-up of 4.8 years (IQR: 2.0-8.9), KRT was initiated in 171 participants (32.1%). Models in group 1 (n = 433) and 2 (n = 533) demonstrated excellent discrimination ability (C-statistic: 0.911-0.972). The calibration slopes exceeded 0.9 across all models, though the Greenwood-Nam-D&apos;Agostino goodness-of-fit test indicated a miscalibration (P < 0.001). Enriched models incorporating the eGFR slope showed the closest alignment with the observed risks. Conclusion: These findings underscore the potential utility of the calculator in improving prognostication and clinical decision-making in pediatric CKD.