{
  "abstract": "Background High recurrence rates following radical hepatectomy for hepatocellular carcinoma (HCC) remain a major clinical challenge. While postoperative adjuvant transcatheter arterial chemoembolization (TACE) is a critical strategy to reduce recurrence risk, its efficacy is highly heterogeneous, and its benefit in specific low-risk populations is controversial. This study aimed to construct a clinical prediction model using the Classification and Regression Tree (CART) algorithm to precisely identify HCC subgroups that derive a survival benefit from adjuvant TACE.Methods A multicenter retrospective cohort study was conducted involving 1,810 HCC patients across four tertiary hospitals in China. Survival differences between TACE and non-TACE groups were initially compared in the overall cohort. A training set (n=603) and internal validation set (n=259) were established for model development. Independent prognostic factors were screened using LASSO regression and multivariate Cox models. A survival decision tree was constructed using the CART algorithm, and its performance was evaluated via the Concordance Index (C-index) and time-dependent ROC curves. Finally, propensity score matching (PSM) was employed within the identified risk strata to verify the model’s guidance value for disease-free survival (DFS).Results In the unselected general population, adjuvant TACE showed no significant improvement in DFS (P=0.840) or overall survival (P=0.160). Multivariate analysis identified five independent predictors of recurrence: irregular tumor morphology, high tumor burden, incomplete tumor capsule, microvascular invasion (MVI), and a neutrophil-to-lymphocyte ratio (NLR) >2.4. The CART model stratified patients into eight risk nodes and classified them into Benefit, Potential Benefit, and Low Benefit groups. PSM analysis demonstrated that adjuvant TACE significantly prolonged DFS in the Benefit group (P=0.036), whereas no survival benefit was observed in the Low Benefit group (P=0.980). The model achieved C-indices of 0.655 and 0.636 in the training and validation sets, respectively.Conclusions The efficacy of adjuvant TACE after hepatectomy is highly population-specific. The CART decision tree model, incorporating tumor morphology, burden, and inflammatory markers, effectively identifies dominant subgroups that benefit from TACE. This provides a precise tool for individualized postoperative management, helping clinicians avoid overtreatment in low-risk patients.",
  "authors": [
    {
      "affiliations": [
        "Department of Gastroenterology, The First Affiliated Hospital of Xi’an Medical University, China"
      ],
      "name": "Ya-Nan Zhang"
    },
    {
      "affiliations": [
        "Department of Gastroenterology, The First Affiliated Hospital of Xi’an Medical University, China"
      ],
      "name": "Ming-Xin Zhang"
    }
  ],
  "title": "IDDF2026-ABS-0205 Machine learning-based prediction of survival benefit from postoperative adjuvant TACE for hepatocellular carcinoma: a multicenter study",
  "uid": "24c9e136-a611-5387-a54d-f27e58ca1d00"
}