Prognostic Analysis of Elderly Patients with Hepatocellular Carcinoma: an Exploration and Machine Learning Model Prediction Based on Age Stratification and Surgical Approach

Chiyu Cai,1,&ast; Hengli Zhu,1,&ast; Bingyao Li,2,&ast; Changqian Tang,3 Yongnian Ren,1 Yuqi Guo,1 Jizhen Li,1 Liancai Wang,1 Deyu Li,1 Dongxiao Li4 1Department of Hepatobiliary and Pancreatic Surgery, Zhengzhou University People’s Hospital, Zhengzhou, Henan, People’s Republic of China; 2Henan Provincial People’s Hospital, Xinxiang Medical University, Zhengzhou, Henan, People’s Republic of China; 3Department of Hepatobiliary and Pancreatic Surgery, Henan University People’s Hospital, Zhengzhou, Henan, People’s Republic of China; 4Department of Gastroenterology, Zhengzhou University People’s Hospital, Zhengzhou, Henan, People’s Republic of China&ast;These authors contributed equally to this workCorrespondence: Deyu Li, Zhengzhou University People’s Hospital, No. 7, Wei Wu Road, Jinshui District, Zhengzhou, Henan, People’s Republic of China, Email [email protected] Dongxiao Li, Zhengzhou University People’s Hospital, No. 7, Wei Wu Road, Jinshui District, Zhengzhou, Henan, People’s Republic of China, Email [email protected]: As the global population ages, precise prognostic tools are needed to optimize postoperative care for elderly hepatocellular carcinoma (HCC) patients. This study established a machine learning-driven predictive model to identify key prognostic determinants and evaluate age/surgical approach impacts, overcoming limitations of traditional statistical methods.Methods: This retrospective study included 252 postoperative HCC patients aged ≥ 65 years (mean age 69.0± 4.3; 68.25% male). Patients were randomly divided into training (70%, n=177) and validation sets (30%, n=75). We evaluated 147 machine learning models to establish the optimal predictive model. Patients were grouped by age (> 75 vs ≤ 75 years) and surgical approach (laparoscopic vs open).Results: The LASSO+RSF model showed strong predictive performance with AUC values of 0.869 and 0.818 in the training and validation sets, respectively. Time-dependent AUCs for 1-, 2- and 3-year survival were 0.874, 0.903, and 0.883 in the training set, and 0.878, 0.882, and 0.915 in the validation set. Key predictors included age-adjusted Charlson index (ACCI, LASSO+RSF synergistic weight (LRSW) =0.160), microvascular invasion (0.111), tumor capsule integrity (0.034), and lymphatic invasion (0.023), while three variables (intraoperative blood loss, tumor margin, WBC) were excluded (LRSW< 0.01). A web-based dynamic nomogram (https://cliniometrics.shinyapps.io/LRSF-GeroHCC/) enabled real-time risk stratification. Patients > 75 years had longer length of stay (16 vs 14 days, P=0.033), higher Clavien-Dindo scores (P=0.014), higher ACCI scores (5.5 vs 4.0, P=0.002), and lower PFS (16.5 vs 24 months, P=0.041). Laparoscopic surgery was associated with longer operative time (202.5 vs 159.0min, P< 0.001), shorter length of stay (14 vs 17days, P< 0.001), and lower Clavien-Dindo scores (P=0.038).Conclusion: The LASSO+RSF model provides validated tools for personalized prognosis management in elderly HCC patients, emphasizing age-adapted surgical strategies and comorbidity-focused perioperative care.Keywords: hepatocellular carcinoma, geriatric oncology, age-adjusted Charlson comorbidity index, machine learning, survival prediction, minimally invasive surgery