Assessment of compensated advanced chronic liver disease based on serum bile acids in chronic hepatitis B patients

Area of Science:

• Hepatology
• Biochemistry
• Medical Informatics

Background:

• Chronic liver disease progression increases decompensation risk.
• Compensated advanced chronic liver disease (cACLD) requires accurate identification.
• Bile acid (BA) profiles may serve as biomarkers for liver fibrosis staging.

Purpose of the Study:

• To develop and validate prediction models for identifying cACLD.
• To utilize individual bile acid (BA) profiles for cACLD detection.
• To apply machine learning algorithms for fibrosis staging in hepatitis B virus (HBV) infection.

Main Methods:

• Prospective recruitment of 159 HBV patients and 60 healthy volunteers.
• Liver stiffness measurement (LSM) categorized patients into F1, F2, and cACLD groups.
• Random Forest (RF) and Support Vector Machine (SVM) models developed using serum BA profiles.

Main Results:

• Serum BA levels, particularly conjugated forms, significantly increased with fibrosis severity.
• Optimized RF model achieved scores of 0.82 for precision, recall, and F1.
• Optimized SVM model achieved scores of 0.86 (precision), 0.84 (recall), and 0.85 (F1).
• RF and SVM models successfully identified key BA features distinguishing cACLD in HBV patients.

Conclusions:

• Individual BA profiles, analyzed via RF and SVM models, offer a novel tool for cACLD identification.
• This approach can assist clinicians in managing chronic liver disease patients.
• Accurate cACLD detection facilitates timely intervention and improves patient outcomes.