ML Predicts Response to TPE in Pediatric ALF: JCEH | July 2026

Introduction:

Therapeutic plasma exchange (TPE) is increasingly used in pediatric acute liver failure (ALF), but it artificially lowers bilirubin and INR, making it difficult to determine whether a child is truly improving or requires urgent liver transplantation. This study developed machine learning (ML) models to predict response to TPE and support real-time clinical decision-making.

Why was this study needed?

• TPE alters conventional prognostic markers, complicating transplant decisions.
• Reliable early predictors of TPE response are lacking.
• Delayed recognition of treatment failure may adversely affect survival.
• Machine learning may improve individualized risk prediction.
• A real-time clinical decision support tool could enhance patient management.

Results:

• Machine learning models accurately predicted response to TPE, with the best performance 12–18 hours after the first TPE session, allowing early identification of responders and non-responders.
• Logistic Regression, Support Vector Machine (SVM), and XGBoost showed the highest predictive performance, consistently outperforming other ML models.
• These validated models were integrated into a web-based clinical decision support tool, enabling real-time prediction of TPE response to assist liver transplant decision-making.

Clinical Impact:

This study demonstrates that artificial intelligence can support critical decision-making in pediatric acute liver failure by identifying children unlikely to respond to TPE at an early stage. Such tools may help optimize transplant timing while avoiding unnecessary delays in life-saving treatment.

Bottom Line:

Machine learning can accurately predict response to therapeutic plasma exchange in pediatric acute liver failure. A simple web-based prediction tool may assist clinicians in determining which children are likely to recover with TPE and which may require urgent liver transplantation.