Dynamic Risk Modelling of Hepatocellular Carcinoma

Despite advancements in the detection and treatment of hepatocellular carcinoma (HCC), the overall survival remains poor. Traditional survival analyses, such as the Cox model, are limited in capturing the dynamic progression across different clinical states. Our paper proposes the utilization of a continuous-time multi-state Markov model to inform risk stratification and management strategies for HCC by accounting for transitions between disease states.

This cohort study included 934 adult patients (25.0% female) with HCC who underwent curative treatment, defined as liver transplantation, resection or ablation, across eight tertiary centres in Australia, China, Japan, Singapore, South Korea and the United States. The primary objective was to assess the risk of HCC recurrence and survival following curative treatment.

The median (IQR) age of the cohort was 65 (IQR 58-74), and 72% had known cirrhosis. Distinct clinical trajectories were identified: recurrence, death without recurrence and death after recurrence. The median (IQR) time from curative treatment to recurrence, from recurrence to death, and from curative treatment to death without recurrence was 15.40 months (4.78-26.01), 51.27 months (20.04-82.50), and 23.13 months (9.26-37.01), respectively. Analyses revealed that advancing age and HCV were associated with recurrence risk, while liver transplantation was protective against recurrence. Ablation, non-curative locoregional therapy, and systemic therapies were associated with higher risks of recurrence and post-recurrence death. Alpha-fetoprotein, tumour size, INR, bilirubin and advanced BCLC stage were key predictors of recurrence and mortality.

By modelling disease as a sequence of interlinked transitions, we provide updated estimates for the time spent within each transition state and predictors for disease progression within a dynamic framework.