{
  "abstract": "Background While the 5-year recurrence rate in patients with early-stage hepatocellular carcinoma (HCC) is around 50-70%, conventional prediction models have mainly focused on tumor characteristics and baseline comorbidities. These models may not account for the contribution of underlying liver diseases, which also play an important role in recurrence. Dynamic laboratory markers could potentially capture changes in liver fibrosis and inflammation, thereby providing more precise recurrence risk stratification.Methods We conducted a nationwide cohort study using Taiwan’s National Health Insurance Laboratory Database (NHILD). Patients with newly diagnosed HCC who received curative treatment between January 1, 2015, and December 31, 2023, were included. We analyzed the longitudinal trajectories of Fibrosis (FIB-4) score, alpha-fetoprotein (AFP), and glycated hemoglobin (HbA1c) from 1.5 years before the curative treatment to the treatment date. These biomarkers were categorized into nine groups based on cutoff values and changes. Additional variables included curative modality, demographic factors, comorbidities, medications exposed after therapy, and the mean Air Quality Index (AQI). SHapley Additive exPlanations (SHAP) were applied to the XGBoost model to assess the feature importance and effect direction.Results We included 31,585 patients with newly diagnosed HCC undergoing curative therapy. Figure 1 ( IDDF2026-ABS-0359 Figure 1. SHAP summary plot of clinical predictors for 5-year hepatocellular carcinoma recurrence) shows the XGBoost SHAP analysis, in which statin use, curative modalities, AQI, HBV antiviral use, and AFP were the five most influential features for recurrence. Post-treatment use of statin, HBV antiviral, metformin, and aspirin had negative SHAP values, indicating lower recurrence risk. In contrast, higher categories of AFP, FIB-4, HbA1c, and AQI were associated with positive SHAP values, indicating greater contributions to predicted recurrence. Curative modality was also an important predictor, with RFA showing positive SHAP values.Conclusions Dynamic changes in laboratory markers, along with post-treatment medication use, curative modality, and AQI, were important contributors to recurrence prediction. These findings demonstrate that longitudinal laboratory measurements are crucial for surveillance and assessment of HCC patients after curative therapy.Abstract IDDF2026-ABS-0359 Figure 1",
  "authors": [
    {
      "affiliations": [
        "National Yang Ming Chiao Tung University, Taiwan"
      ],
      "name": "Yen-Ling Chen"
    },
    {
      "affiliations": [
        "National Yang Ming Chiao Tung University, Taiwan"
      ],
      "name": "Chun-Ying Wu"
    }
  ],
  "title": "IDDF2026-ABS-0359 Longitudinal changes in laboratory markers for predicting hepatocellular carcinoma recurrence following curative therapy: a nationwide cohort study",
  "uid": "fdb00fd4-7232-5347-9b11-5e9489af4c94"
}