Validation of the children international IgA nephropathy prediction tool based on data in Southwest China

BACKGROUND: Immunoglobulin A nephropathy (IgAN) is one of the most common kidney diseases leading to renal injury. Of pediatric cases, 25%–30% progress into end-stage kidney disease (ESKD) in 20–25 years. Therefore, predicting and intervening in IgAN at an early stage is crucial. The purpose of this study was to validate the availability of an international predictive tool for childhood IgAN in a cohort of children with IgAN treated at a regional medical centre. METHODS: An external validation cohort of children with IgAN from medical centers in Southwest China was formed to validate the predictive performance of the two full models with and without race differences by comparing four measures: area under the curve (AUC), the regression coefficient of linear prediction (PI), survival analysis curves for different risk groups, and R(2)D. RESULTS: A total of 210 Chinese children, including 129 males, with an overall mean age of 9.43 ± 2.71 years, were incorporated from this regional medical center. In total, 11.43% (24/210) of patients achieved an outcome with a GFR decrease of more than 30% or reached ESKD. The AUC of the full model with race was 0.685 (95% CI: 0.570–0.800) and the AUC of the full model without race was 0.640 (95% CI: 0.517–0.764). The PI of the full model with race and without race was 0.816 (SE = 0.006, P < 0.001) and 0.751 (SE = 0.005, P < 0.001), respectively. The results of the survival curve analysis suggested the two models could not well distinguish between the low-risk and high-risk groups (P = 0.359 and P = 0.452), respectively, no matter the race difference. The evaluation of model fit for the full model with race was 66.5% and without race was 56.2%. CONCLUSIONS: The international IgAN prediction tool has risk factors chosen based on adult data, and the validation cohort did not fully align with the derivation cohort in terms of demographic characteristics, clinical baseline levels, and pathological presentation, so the tool may not be highly applicable to children. We need to build IgAN prediction models that are more applicable to Chinese children based on their particular data.