273 - Redefining Breast Cancer Tumor Size Classification through Big Data Analytics: A New Approach to Enhancing Prognostic Accuracy and Personalized Treatment

B. Feng, X. Yang, H. Luo, L. Chen, L. Tan, and F. Jin; Radiation Physics Center, Chongqing University Cancer Hospital, Chongqing, China

Purpose/Objective(s): The current definitions of breast cancer tumor sizes by the American Joint Committee on Cancer (AJCC) are primarily based on empirical, human-defined categories. This approach may not fully capture the complexity of tumor biology and its implications for patient outcomes. With the advancement of big data technologies, we have the opportunity to understand the relationship between tumor size and breast cancer prognosis from a more nuanced perspective. This study aims to leverage big data analytics to redefine the AJCC classification of breast cancer tumor sizes, in hopes of discovering more precise tumor size staging criteria. This could provide more personalized treatment recommendations and prognostic assessments for breast cancer patients. Materials/

Methods: Utilizing data from the Surveillance, Epidemiology, and End Results (SEER) program (2004-2015), we enrolled 88,560 breast cancer patients, focusing on those categorized within T1-T4, N0, and M0 stages, and 35,515 breast cancer patients from 2018-2020 SEER program were used for validating our method. Hierarchical clustering was employed to categorize tumors into distinct stages based on size and patient survival data. we developed clustering models for both three-category and four-category classifications. To validate the effectiveness of these models, we employed the encompassing Kaplan-Meier survival analysis.
Results: These stages are delineated as follows: Stage I encompasses tumor sizes from 1 to 14 mm[5-year survival rate(SR): 0.91, 10-year SR: 0.78, 15-year SR: 0.64], Stage II from 15 to 34 mm(5-year SR: 0.84, 10-year SR: 0.69, 15-year SR: 0.56), Stage III from 35 to 120 mm(5-year SR: 0.73, 10-year SR: 0.57, 15-year SR: 0.47), and Stage IV represents tumors exceeding 120 mm(5-year SR: 0.63, 10-year SR: 0.44, 15-year SR: 0.35) in size. Compared to the traditional staging system, the Log-rank test revealed significant differences between survival curves (P < 0.05), indicating that the new staging system more objectively differences in patient prognosis.
Conclusion: This study underscores the value of big data in refining oncological practices and opens avenues for further research into the optimization of tumor staging criteria.