Hepatocellular carcinoma (HCC) recurrence after treatment poses a significant burden on patients and their families, particularly in elderly cases (≥ 65 years old) combining type 2 diabetes mellitus (T2DM). T2DM is known to exacerbate the risk of HCC recurrence due to mechanisms such as insulin resistance, oxidative stress, and chronic inflammation. Despite advances in HCC management, including transarterial chemoembolization (TACE), recurrence rates remain high, necessitating better predictive models. Survival and recurrence outcomes of HBV-related elderly HCC patients with T2DM undergoing TACE were retrospectively analyzed. Clinical data, including tumor size, tumor number, gamma-glutamyl transferase (GGT), and other laboratory parameters until recurrence free survival (RFS), were collected. Machine learning techniques, including random survival forest (RSF) for variable selection, were employed to develop a predictive nomogram. The model’s performance was evaluated using receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA) in training and validation cohorts. An external validation cohort consisting of 97 patients from an independent medical center was additionally included to assess model generalizability. A total of 426 elderly patients with HBV-related HCC and T2DM were enrolled, with a mean age of 69.2 years and a male predominance. Tumor size, tumor number, and GGT were identified as key risk factors for HCC recurrence in these patients. The nomogram demonstrated strong predictive capability for 1-year, 3-year, and 5-year RFS, with area under the curve (AUC) values ranging from 0.616 to 0.696 across cohorts. Calibration and DCA confirmed the model’s accuracy and clinical applicability. Patients classified as high-risk by the model exhibited significantly shorter survival compared to low-risk patients in both training and validation cohorts. The nomogram demonstrated favorable discrimination and calibration in both internal and external validation cohorts, indicating good consistency across datasets. This study highlights the importance of tumor size, tumor number, and GGT as critical predictors of HCC recurrence in HBV-related patients with T2DM. The proposed machine learning-based nomogram offers a practical tool for personalized risk assessment, enabling clinicians to optimize treatment strategies and improve outcomes in this high-risk population.
Liu et al. (2026) studied this question.