Background and objectives: This study aimed to identify key prognostic factors influencing survival in patients with non-metastatic breast cancer. A predictive model was developed using the Random Survival Forest (RSF) method to enhance survival estimation and support clinical decision-making.
Materials and Methods: In this retrospective cohort study, the medical records of 767 patients with non-metastatic breast cancer who were treated at the Mahdia Radiotherapy Center in Hamadan between 2006 and 2018 were reviewed. After excluding incomplete records, 442 patients remained for the final analysis. Demographic, clinical, and treatment-related data were extracted from the patients’ medical charts. Both the Cox proportional hazards (PH) regression (Cox regression) model and the RSF model were applied to identify significant predictors of survival.
Results: The mean age of the patients was 49.23 ± 10.85 years, and the mean survival time was 21.09 ± 2.27 months. One-, three-, and five-year survival rates were 98.8%, 97.5%, and 97.2%, respectively. Based on the RSF model, radiotherapy dose, recurrence, age, nodal stage (N), and overall stage were identified as the most influential predictors of survival. HER2 status, initial treatment approach, and surgical method were also included in the model. The RSF model achieved a C-index ranging from 0.63 to 0.73, outperforming the Cox regression model (C-index 0.54–0.66).
Conclusion: The RSF model effectively identified important predictors of survival in non-metastatic breast cancer and may provide a valuable tool for personalized clinical decision-making. These findings demonstrate the potential value of machine learning-based models in oncology research and patient management.