Prognostic factors and a Cox proportional hazards model for survival in primary liver cancer

Aim: Accurate prognostic assessment using biochemical indices is essential for personalized management of primary liver cancer in community-level hospitals. This study aimed to construct and validate a Cox regression-based nomogram integrating these indices with clinicopathological data to predict overall survival (OS). Methods: In this retrospective study, 1,124 patients with primary liver cancer were randomly allocated in a 7:3 proportion into training and validation cohorts. Univariable and multivariate Cox proportional hazards analyses were performed to identify independent prognostic factors and establish a nomogram for predicting OS at 1, 3, and 5 years. The model’s discrimination was assessed via the concordance index (C-index) and time-dependent receiver operating characteristic (ROC) curves. Furthermore, calibration curves and decision curve analysis (DCA) were utilized to evaluate predictive agreement and clinical utility. Results: Multivariate analysis identified an elevated aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio, high alpha-fetoprotein (AFP ≥ 400 ng/mL), reduced body mass index (BMI < 18.5 kg/m2), and advanced tumor-node-metastasis (TNM) stage as independent adverse predictors for OS, while surgical resection and hepatitis history favored survival. The nomogram showed moderate discrimination, yielding C-indices of 0.678 (training cohort) and 0.659 (validation cohort), with corresponding areas under the curve (AUCs) for years 1, 3, and 5 of 0.725, 0.733, and 0.779 in the training cohort and 0.700, 0.718, and 0.750 in the validation cohort. Furthermore, the model effectively stratified patients into distinct prognostic risk groups (log-rank test P < 0.001). Conclusion: This nomogram may assist clinicians in risk stratification and survival estimation for primary liver cancer patients in primary-level healthcare settings.