#7 The kidney failure risk equation in IgA nephropathy: external validation and model update

Abstract Background and Aims IgA nephropathy (IgAN) is the most prevalent glomerulonephritis and a leading cause of chronic kidney disease (CKD) globally. The disease's progression varies widely, so accurately predicting prognosis is crucial for identifying patients at risk of developing end-stage kidney disease (ESKD). Prognostic information is essential in patient counseling and could help decide appropriate treatment options. The 2024 KDIGO guidelines for CKD recommend using established CKD risk prediction tools to assess the risk for ESKD. The Kidney Failure Risk Equation (KFRE), developed in 2011 from a Canadian cohort, predicts the 2- and 5-year risk of ESKD in patients with eGFR <60 ml/min/1.73 m2 corresponding to chronic kidney failure stage 3. It has been externally validated using multiple international CKD cohorts involving over 700,000 patients. The tool includes four variables: age, sex, urine albumin/creatinine, and eGFR. It is easily accessible as an online tool. The International IgA Prediction Tool was developed in 2019. It can predict prognosis in IgAN for up to 6.7 years, and the current KDIGO guidelines recommend using it in patient counseling. However, the tool needs clinical and histopathological features from the Oxford classification to predict prognosis. In many patients with IgAN, detailed histopathological information might be unavailable due to few glomeruli in the diagnostic kidney biopsy. We, therefore, set out to externally validate the KFRE in an IgAN population. Method We used data from the Norwegian Kidney Biopsy Registry and patient records to identify 236 patients with biopsy-confirmed IgAN who had advanced to stage 3 chronic kidney disease. We used the published regression equation to derive the five-year prognostic index. We then assessed discrimination using cumulative dynamic receiver operating characteristics (ROC) analysis and the concordance index. Model calibration was evaluated by calibration curves, while goodness of fit was assessed by the Akaike information criterion (AIC). Recalibration was performed by updating the baseline survival from the validation cohort and performing regression on the prognostic index. An updated multivariable Cox model was derived using the same four variables and internally validated using boot-strapping methodology. Results 170 (72%) of the patients were male, and the median age at CKD stage 3 was 48 years (IQR 34–59). The median urine albumin creatinine ratio was 71 mg/mmol (IQR 28–216). A total of 98 patients reached ESKD during the time of follow-up, while 20 patients died. The median follow-up time was four years (IQR 1–9). In total, 167 (71%) patients were treated with RAAS inhibitors, while 57 (24%) received immunosuppressive treatment. ROC analysis at five years presented an area under curve (AUC) value at 0.78, decreasing to 0.64 at 20 years (Table 1). Calibration curves revealed poor calibration at five and ten years by underestimating and overestimating ESKD-free survival in the low-risk and high-risk groups, respectively (Fig. 1). The AIC value was 860.61. Recalibration showed an improved calibration after five and ten years and improved goodness of fit (AIC: 838.21). Multivariable analysis revealed that eGFR and proteinuria were the prominent predictors. The new Cox model showed improved discrimination compared to the KRFE model, with AUC at 0.84 at five years, decreasing to 0.73 at 20 years (Table 1), and no signs of overfitting in internal validation, improved calibration (Fig. 1), and an improved AIC at 815.093. Conclusion Based on data from this cohort, the 4-parameter KRFE has acceptable predictive probabilities in patients with IgAN. However, it might be possible to improve model performance if the KFRE is adapted and updated for IgAN patients. One should be aware of the potential reduced predictive performance of generic CKD calculators in subgroups of patients.