A Machine Learning Model to Predict De Novo Hepatocellular Carcinoma Beyond Year 5 of Antiviral Therapy in Patients With Chronic Hepatitis B

  • 0Department of Gastroenterology, CHA Bundang Medical Center, CHA University, Seongnam-si, Gyeonggi-do, South Korea.

Summary

This summary is machine-generated.

This study developed a machine learning model to predict hepatocellular carcinoma (HCC) risk in chronic hepatitis B (CHB) patients on antiviral therapy. The model accurately identifies high-risk individuals for personalized surveillance.

Area Of Science

  • Hepatology
  • Machine Learning in Medicine
  • Oncology

Background

  • Chronic hepatitis B (CHB) patients on long-term antiviral therapy require risk stratification for hepatocellular carcinoma (HCC).
  • Predicting HCC development beyond 5 years of entecavir (ETV) or tenofovir (TFV) therapy is crucial for patient management.

Purpose Of The Study

  • To develop and validate a machine learning (ML) model for predicting HCC risk in CHB patients after 5 years of ETV/TFV therapy.
  • To create a tool for individualized HCC surveillance strategies.

Main Methods

  • Utilized data from CHB patients treated with ETV/TFV for over 5 years, excluding those diagnosed with HCC in the first 5 years.
  • Developed models using 36 variables including baseline characteristics and laboratory values at baseline, 5 years, and changes over time.
  • Applied and validated five ML algorithms, including logistic regression and random forest, through internal and external validation.

Main Results

  • An ensemble ML model combining logistic regression and random forest demonstrated superior performance.
  • The ensemble model achieved an AUC of 0.811 (balanced accuracy 0.754) in the training set and 0.862 (balanced accuracy 0.771) in the external validation cohort.
  • A web-based calculator was developed based on the validated model.

Conclusions

  • The developed ML model accurately predicts HCC risk in CHB patients beyond 5 years of antiviral therapy.
  • The model facilitates individualized HCC surveillance by stratifying patient risk.
  • This tool can aid clinicians in optimizing monitoring strategies for CHB patients at risk of HCC.