Validation of 2 prognostic models to predict renal allograft outcome after IgA nephropathy recurrence

Abstract

Introduction: IgA nephropathy (IgAN) recurrence (IgANr) after kidney transplantation (KTx) is common and contributes to reducing graft survival. Some tools have been developed to predict the patients who are at a higher risk of poor outcomes among the native (international IgAN prediction tool [IIgAN-PT]) and graft (Bednarova's prediction tool [Bednarova-PT]) kidney. We aimed to analyze their performance in a KTx population other than the originally reported. Methods: We performed a multicenter retrospective study including KTx with biopsy-proven IgANr. IIgAN-PT and Bednarova-PT were used to calculate the risk of death-censored graft loss (DCGL). We assessed the performance of both prediction models using discrimination and calibration metrics and Kaplan-Meier plots. Results: One hundred twenty KTx with IgANr were included. The time-dependent receiver operating characteristic (ROC) area under the curve (AUC) of Bednarova-PT for predicting DCGL was 83.5 (95% CI: 72.3-94.7) and the calibration slope was 0.96 (95% CI: 0.37-1.49). The time-dependent ROC AUC of IIgAN-PT for predicting DCGL was 87.3 (95% CI: 77.58-97.02) and the calibration slope was 2.49 (95% CI: 0.19-4.13). IIgAN-PT tended to underestimate the graft-loss risk in high-risk individuals. The Kaplan-Meier curve of the highest risk group, defined by using both prediction tools, was clearly separated from the other curves. Conclusion: Both IIgAN-PT and Bednarova-PT performed well in predicting DCGL after IgANr and should be used to identify those KTx at the highest risk. Both models had good discriminatory ability and were well-calibrated, although the calibration slope was higher for IIgAN-PT, tending to underestimate the risk in high-risk individuals.