Prognostic nomograms for hormone receptor-positive breast cancer with lung metastasis: a SEER-based study

Background: The paucity of validated prognostic tools for breast cancer with lung metastasis (BCLM), compounded by uncharacterized molecular subtype-specific survival patterns, underscores critical unmet needs in metastatic breast cancer management. Leveraging the Surveillance, Epidemiology, and End Results (SEER) registry's extensive dataset, this study pioneers a machine learning-enhanced nomogram that simultaneously predicts lung metastasis-related mortality risks in Hormone receptor (HR)-positive patients and generates personalized survival trajectories, thereby enabling molecularly stratified clinical decision-making. Methods: From the SEER database (2010-2015), 1,807 eligible BCLM cases underwent 7:3 random split into training cohorts (n=1,264) and validation cohorts (n=543), with 2016 cases (n=374) as external cohort. Multivariable Cox models identified independent mortality determinants for both overall survival (OS) and breast cancer-specific survival (BCSS), informing the development of a 1-3 year prognostic nomogram. The prognostic accuracy of the nomogram underwent tripartite validation: calibration curves, receiver operating characteristic (ROC) curves, the Harrell’s concordance index (C-index) and decision curve analysis (DCA). Results: Univariate screening followed by multivariable Cox proportional hazards regression identified 11 independent prognostic determinants spanning three domains: (1) demographic (age, race, marital status), (2) tumor biology (grade, T stage, HER2 status, hepatic/bony/cerebral metastases), and (3) therapeutic interventions (surgery, chemotherapy), all significantly associated with OS/BCSS (P<0.05). Subsequent development of 12-/24-/36-month survival nomograms demonstrated exceptional temporal validity. The training cohort showed C-indexes of 0.655 (95% CI, 0.637-0.673) for OS and 0.662 (95% CI, 0.644-0.680) for BCSS, while in the validation cohort, the nomogram achieved C-indexes of 0.648 (95% CI, 0.621-0.675) for OS and 0.657 (95% CI, 0.630-0.684) for BCSS.ROC curves, calibration plots, and DCA results indicated strong predictive performance of nomograms. Conclusion: The validated prognostic nomograms demonstrated robust predictive accuracy for BCLM outcomes, offering clinicians actionable tools to facilitate personalized surveillance and evidence-based therapeutic interventions.