IgA nephropathy (IgAN) is the most prevalent primary glomerular disease worldwide and a leading cause of end-stage kidney disease (ESKD). Its clinical heterogeneity results in divergent renal outcomes, making early identification of high-risk patients essential. Prognostic models are crucial for stratifying ESKD risk, guiding treatment intensity, optimizing timing of interventions such as immunosuppressive therapy, and informing clinical trial enrollment. Over recent decades, multiple prognostic approaches have emerged, ranging from traditional clinical and histopathological scoring systems to advanced machine learning (ML) and deep learning (DL) models designed to capture complex nonlinear interactions and improve predictive precision. Among them, the International IgA Nephropathy Prediction Tool (IIgAN-PT), endorsed by the 2021 KDIGO guidelines, represents a landmark in globally validated risk assessment and has set the foundation for standardized clinical decision support. However, classical models often rely on static baseline parameters and may not adequately reflect dynamic disease trajectories, limiting their utility in real-time clinical management. To overcome these limitations, ML- and DL-based models increasingly integrate multi-omics data, serial clinical measurements, and digital pathology features, offering enhanced accuracy, dynamic risk tracking, and potential for personalized response prediction. These data-driven approaches are progressively bridging the gap between prognostic research and precision nephrology. This review provides a comprehensive overview of the evolution of IgAN prognostic models, summarizes their strengths and limitations, and discusses considerations for clinical translation. By highlighting emerging trends toward explainable AI, dynamic time-series modeling, and multimodal prognostication, we outline how next-generation prediction tools may enable real-time, AI-driven decision support for individualized IgAN management.
Xu et al. (2026) studied this question.