Analysis of Prognostic Factors and Nomogram Construction for Postoperati i ve Survival of Triple-Negative Breast Cancer

Chenxi Wang ¹ Xiangqian Zhao ² Dawei Wang ¹ Jinyun Wu ¹ Jizhen Lin ² Weiwei Huang ^3* Yangkun Shen ¹ Qi Chen ¹

• ¹ Fujian Key Laboratory of Innate Immune Biology, Biomedical Research Center of South China, College of Life Science, Fujian Normal University, Fuzhou, Fujian, 350117, Fuzhou, China
• ² The Cancer Center, Fujian Medical University Union Hospital, No.29, Xinquan Road, Gulou District, Fuzhou, Fujian, 350001, Fuzhou, China
• ³ Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital, Fujian Provincial Key Laboratory of Translational Cancer Medicine,, Fuzhou, China

Triple-negative breast cancer (TNBC) is a highly aggressive breast cancer subtype associated with poor prognosis and limited treatment options. This study utilized the SEER database to investigate clinicopathologic characteristics and prognostic factors in TNBC patients. Machine learning algorithms specifically Gradient Boosting Machines (XGBoost) and Random Forest classifiers were applied to develop survival prediction models and identify key prognostic markers. Results indicated significant predictors of survival, including tumor size, lymph node involvement, and distant metastases. Our proposed work showed better predictive performance, with a C-index of 0.8544 and AUC-ROC values of 0.9008 and 0.8344 for one year and three year overall survival predictions. Major predictors of survival comprises tumor size, HR is 3.657 for T4, lymph node involvement, HR is 3.018 for N3, distant metastases, HR is 1.743 for M1, and prior treatments includes surgery, HR is 0.298, chemotherapy, HR is 0.442, and radiotherapy, HR is 0.607. The findings emphasize the clinical utility of AI-driven models in improving TNBC prognosis and guiding personalized treatment strategies. This study provides novel insights into the survival dynamics of TNBC patients and underscores the potential of predictive analytics in oncology.