Predictive models of chronic kidney disease progression in pediatric patients

We read with great interest the study conducted by Ng et al., 1 Ng D.K. Matheson M.B. Schwartz G.J. et al. Development of an adaptive clinical web-based prediction tool for kidney replacement therapy in children with chronic kidney disease. Kidney Int. 2023; 104: 985-994 Abstract Full Text Full Text PDF Scopus (0) Google Scholar aimed to develop a suite of predictive models for time to kidney replacement therapy in children with chronic kidney disease (CKD). In this well-designed study, Ng et al.1 Ng D.K. Matheson M.B. Schwartz G.J. et al. Development of an adaptive clinical web-based prediction tool for kidney replacement therapy in children with chronic kidney disease. Kidney Int. 2023; 104: 985-994 Abstract Full Text Full Text PDF Scopus (0) Google Scholar used robust data from the Chronic Kidney Disease in Children study 2 Furth S.L. Cole S.R. Moxey-Mims M. et al. Design and methods of the Chronic Kidney Disease in Children (CKiD) prospective cohort study. Clin J Am Soc Nephrol. 2006; 1: 1006-1015 Crossref PubMed Scopus (339) Google Scholar and a combination of sophisticated strategies, including both conventional statistics and machine learning methods. The authors developed 6 models of CKD progression in pediatric patients. In external validation, the elementary model, which includes the glomerular filtration rate, urine protein-creatinine ratio, and the CKD cause, showed excellent discrimination and calibration. Interestingly, we previously reported a risk prediction model for kidney failure based on a cohort of 147 pediatric patients enrolled in our Predialysis Interdisciplinary Management Program. 3 Cerqueira D.C. Soares C.M. Silva V.R. et al. A predictive model of progression of CKD to ESRD in a predialysis pediatric interdisciplinary program. Clin J Am Soc Nephrol. 2014; 9: 728-735 Crossref PubMed Scopus (32) Google Scholar Similarly, using classic survival analysis, we found that the most accurate model included baseline kidney function, proteinuria at admission, and primary kidney disease (glomerular vs. nonglomerular disease). Moreover, the c-statistic of our model was 0.872 (95% confidence interval, 0.802–0.942) for the 5-year follow-up, similar to that of the elementary model reported by Ng et al. (c-statistic = 0.865). Taken together, these findings highlight that routinely available clinical and laboratory data (namely, underlying CKD cause, glomerular filtration rate values at the early stages of disease, and proteinuria) are strong predictors of CKD progression. 4 Crane C.R. Garimella P.S. Heinze G. Predicting pediatric kidney disease progression-are 3 variables all you need?. Kidney Int. 2023; 104: 885-887 Abstract Full Text Full Text PDF Google Scholar We emphasize that these models are based on clinically accessible data. Therefore, the user-friendly web-based tool proposed by Ng et al.1 Ng D.K. Matheson M.B. Schwartz G.J. et al. Development of an adaptive clinical web-based prediction tool for kidney replacement therapy in children with chronic kidney disease. Kidney Int. 2023; 104: 985-994 Abstract Full Text Full Text PDF Scopus (0) Google Scholar has great potential to help pediatric nephrologists and CKD clinics in clinical decision-making and patient care planning. Development of an adaptive clinical web-based prediction tool for kidney replacement therapy in children with chronic kidney diseaseKidney InternationalVol. 104Issue 5PreviewClinicians need improved prediction models to estimate time to kidney replacement therapy (KRT) for children with chronic kidney disease (CKD). Here, we aimed to develop and validate a prediction tool based on common clinical variables for time to KRT in children using statistical learning methods and design a corresponding online calculator for clinical use. Among 890 children with CKD in the Chronic Kidney Disease in Children (CKiD) study, 172 variables related to sociodemographics, kidney/cardiovascular health, and therapy use, including longitudinal changes over one year were evaluated as candidate predictors in a random survival forest for time to KRT. Full-Text PDF