Conventional machine learning-based prediction models did not outperform the International IgA Nephropathy Prediction...

Area of Science:

• Nephrology
• Medical Informatics
• Biostatistics

Background:

• Immunoglobulin A nephropathy (IgAN) is a leading cause of end-stage kidney disease (ESKD).
• The International IgA Nephropathy Prediction Tool (IIgAN-PT) currently predicts IgAN prognosis.
• There is a need to improve prediction performance using machine learning (ML) methods.

Purpose of the Study:

• To develop and evaluate ML-based models for predicting kidney disease progression in IgAN patients.
• To compare the performance of ML models against the existing IIgAN-PT.

Main Methods:

• Analysis of 4,425 biopsy-confirmed IgAN patients from nine Korean hospitals.
• Development of four ML models: CatBoost, optimized CatBoost with Cox, deep Cox hazards, and deep Cox mixture models.
• Evaluation of model performance using Area Under the Curve (AUC) and calibration plots, compared to IIgAN-PT.

Main Results:

• The IIgAN-PT full model demonstrated excellent performance (AUC 0.896 for 5-year outcome).
• ML-based models showed good predictive performance in external validation (AUCs ranging from 0.823 to 0.847).
• ML models slightly underestimated risks in the external validation cohort; overall performance was non-inferior to IIgAN-PT.

Conclusions:

• ML-based models demonstrate good performance in predicting adverse kidney outcomes in IgAN.
• The developed ML models did not outperform the established IIgAN-PT in this study.
• Further research may be needed to refine ML approaches for IgAN prognosis.