ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: TH-PO0739

Utility of Conventional Pathologic Scoring Techniques in Predicting Glomerular Disease Progression Using Machine Learning

Session Information

Category: Glomerular Diseases

  • 1402 Glomerular Diseases: Clinical, Outcomes, and Therapeutics

Authors

  • Bou Matar, Raed, Cleveland Clinic, Cleveland, Ohio, United States
  • Banker, Margaret, Northwestern University, Evanston, Illinois, United States
  • Smith, Abigail R., Northwestern University, Evanston, Illinois, United States
  • Nast, Cynthia C., Cedars-Sinai Medical Center, Los Angeles, California, United States

Group or Team Name

  • The CureGN Consortium.
Background

Predicting glomerular disease progression is critical for optimizing treatment strategies, yet the reliability of conventional pathology scores in forecasting clinical outcomes remains uncertain.

Methods

This study included children and adults from the CureGN, a multi-center observational cohort. Conventional morphologic pathology features were centrally semiquantitatively scored and assessed for reliability. These pathology features and basic demographics were used as input features to train and cross-validate machine learning (ML) models predicting clinical outcomes. ML models included K-nearest neighbors (KNN), Random Forest (RF), Support Vector Machines (SVM), Extreme Gradient Boosting (XGB), and Random Survival Forest (RSF), with hyperparameter tuning via Sci-kit Learn and XGBoost. Predictions included progression to a composite endpoint of ESKD or 40% eGFR decline within 4 y of biopsy, time from enrollment to remission, and to the composite endpoint. Model performance was compared using area under the curve (AUC) or integrated AUC (iAUC), average precision (AP), Brier score, and concordance index (c-index) as appropriate. Feature importance was assessed using Shapley Additive explanations (SHAP).

Results

The analysis included 1166 patients (236 minimal change disease, 273 focal segmental glomerulosclerosis, 393 IgA nephropathy and 264 membranous nephropathy). Figure 1 illustrates performance of various models to predict outcomes at 4 y post-study enrollment, highlighting RF’s superior AP and AUC values. RSF models achieved similar performance in predicting time to composite outcome (iAUC 0.79, c-index 0.74, Brier score 0.12). The most predictive pathology features included interstitial fibrosis and tubular atrophy (IFTA), tubular microcystic changes, and segmental sclerosis lesions, as determined via SHAP analysis.

Conclusion

ML models leveraging conventional pathology features show promising accuracy in predicting clinical outcomes, potentially enhancing personalized treatment strategies.

Funding

  • NIDDK Support

Digital Object Identifier (DOI)