Modeling 5-Year Hepatocellular Carcinoma Risk in Alaska Native Peoples With Hepatitis B Virus Infection.

Area of Science:

• Hepatology and Viral Hepatitis Research
• Cancer Epidemiology and Prevention
• Biostatistics and Predictive Modeling

Background:

• Hepatocellular carcinoma (HCC) risk stratification in Alaska Native (AN) peoples with chronic hepatitis B virus (HBV) infection is crucial.
• Existing HCC risk models lack prediction within 5 years and do not incorporate HBV genotype (GT).

Purpose of the Study:

• To develop an improved HCC risk prediction model for AN peoples with chronic HBV infection.
• To address limitations of existing models by incorporating longitudinal data and HBV genotype.

Main Methods:

• Utilized longitudinal data from 1163 AN peoples with HBV.
• Incorporated predictors including age, sex, GT, alpha-fetoprotein (AFP), and other clinical markers.
• Employed generalized estimation equation and Random Forest approaches for 5-year HCC risk modeling.

Main Results:

• AFP, platelet count, and GT were the most significant predictors.
• The Random Forest model achieved an AUC of 0.88, outperforming existing models.
• Achieved 70% sensitivity at 90% specificity for 5-year HCC risk prediction.

Conclusions:

• Developed a novel HCC risk prediction model using longitudinal patient data.
• The model accurately estimates HCC risk over time, aiding surveillance and stratification.
• The Random Forest model demonstrates superior performance in predicting HCC risk.