Objective:

To investigate the association of primary tumor resection (PTR) with survival outcomes in stage IV breast cancer and develop a model to identify patient characteristics linked to better prognosis following PTR.

Approach:

• Data Source: Utilized the SEER registry (2000-2020) for a propensity-score matched analysis of stage IV breast cancer patients.
• Statistical Methods: Employed Cox regression and Kaplan-Meier methods to estimate overall survival (OS) and cancer-specific survival (CSS).
• Machine Learning Model: Developed and validated a machine learning model for predicting survival outcomes, with a user-friendly web platform for clinical use.

Key Findings:

• PTR was positively associated with overall survival (OS) (HR, 0.61; 95% CI, 0.57 to 0.66) and cancer-specific survival (CSS) (HR, 0.64; 95% CI, 0.59 to 0.67) in stage IV breast cancer patients.
• Subgroup analysis indicated better survival for patients with tumors ≤5 cm (OS, HR, 0.52; 95% CI, 0.44 - 0.63; CSS, HR, 0.52; 95% CI, 0.43 - 0.63), N2 status, and HER2 overexpression (OS, HR, 0.56; 95% CI, 0.52 - 0.61; CSS, HR, 0.56; 95% CI, 0.46 - 0.67).
• The best outcomes were observed with trimodality therapy combining PTR, chemotherapy, and radiotherapy (OS, HR, 0.40; 95% CI, 0.37 - 0.44; CSS, HR, 0.03; 95% CI, 0.01 - 0.10).
• The Support Vector Machine (SVM) model demonstrated high accuracy in survival prediction across external datasets.

Interpretation:

Patients with specific tumor characteristics (≤5 cm, N2 status, HER2 overexpression) showed improved survival after PTR, and the developed ML model can assist in identifying suitable surgical candidates.

Limitations:

• The study focused exclusively on female patients, which may limit generalizability.
• Data was sourced from a single registry, potentially introducing selection bias.

Conclusion:

The ML model based on eight clinical indicators predicts survival and aids in identifying surgical candidates for stage IV breast cancer.