The current guidelines on hepatocellular carcinoma (HCC) aimed to build effective prognostic stratification strategies to guide therapeutic allocation; however, the current guidelines did not consider the simultaneous comparison of distinct therapies in similar populations. Here, the investigators aimed to develop and validate a new, integrated prognostic scheme for HCC patients using artificial intelligence (AI) to simulate the survival outcomes of patients allocated to different treatments.

Given that liver resection (LR) and transarterial chemoembolization (TACE) are the mainstay curative and palliative therapies for HCC, respectively, patients who underwent LR or TACE were included in the study. Various prognostic AI algorithms were modeled using data from a large multi-institutional cohort, where LR and TACE were considered independent factors. The C-index, Brier score (BS), and area under the receiver operating characteristic curve (auROC) were calculated to estimate the AI models.

Inclusion Criteria

• Diagnosed with HCC
• Receiving LR or TACE therapy
• Complete clinical information
• Preserved liver function (Child-Pugh Score [CPS] ≤ 7)
• PS 0-1

Exclusion Criteria

• Vascular invasion or extrahepatic spread;
• Ecompensated liver cirrhosis ;
• Younger than 18 years;
• Duration of follow-up of fewer than 30 days.

Between January 2010 and October 2021, a total of 4991 consecutive HCC patients receiving LR or TACE from 15 Chinese tertiary hospitals were retrospectively screened. HCC was diagnosed based on either imaging or pathological results as previously described