Development and validation of a machine learning model to evaluate survival in patients with newly diagnosed breast cancer with liver metastasis
Authors:
Open Access
- 16 June 2026
- journal article
- research article
- Published by Frontiers Media SA in Frontiers in Oncology
- Vol. 16, 1777850
Abstract
Background: Breast cancer liver metastasis (BCLM) is associated with substantial prognostic heterogeneity and poor survival outcomes. Existing staging systems and prognostic models have a limited capacity for individualized survival prediction for patients with newly diagnosed BCLM, highlighting the need for more accurate and clinically applicable predictive tools. This study aimed to develop a prognostic nomogram for patients with newly diagnosed breast cancer and liver metastases using Surveillance, Epidemiology, and End Results (SEER) database data (2010–2021). Methods: Ten stable prognostic variables—age; tumor size; brain, bone, and lung metastases; histological grade; chemotherapy; estrogen receptor (ER) status; progesterone receptor (PR) status; and human epidermal growth factor receptor-2 (HER2) status—were selected using four machine learning algorithms (Least Absolute Shrinkage And Selection Operator, Extreme Gradient Boosting, Decision Tree, and Random Forest). The SEER cohort was randomly divided into training (70%) and internal validation (30%) cohorts, with an independent external cohort for validation. These variables were incorporated into a multivariate Cox proportional hazards regression model to construct a nomogram. Model performance was evaluated using the concordance index (C-index), calibration curves, decision curve analysis (DCA), and area under the curve (AUC). Results: The C-indices for the training, internal validation, and external validation cohorts were 0.760, 0.740, and 0.787, respectively. The 1-, 3-, and 5-year AUCs were 0.777, 0.757, and 0.764 (training); 0.755, 0.769, and 0.754 (internal validation); and 0.727, 0.752, and 0.801 (external validation), respectively. The calibration curves indicated good agreement and DCA confirmed the clinical utility of the model. Conclusion: The proposed nomogram demonstrated robust predictive performance for patients with breast cancer and liver metastases and may provide valuable prognostic information, pending further validation.Keywords
This publication has 24 references indexed in Scilit:
- A network dynamic nomogram for predicting overall survival and cancer-specific survival in patients with breast cancer liver metastases: an analysis based on the SEER databaseDiscover Oncology, 2024
- Assessment of nuclear grade-based recurrence risk classification in patients with hormone receptor-positive, human epidermal growth factor receptor 2-negative, node-positive high-risk early breast cancerBreast Cancer, 2023
- Prognostic nomogram and risk factors for predicting survival in patients with pT2N0M0 esophageal squamous carcinomaScientific Reports, 2023
- The current staging and classification systems of breast cancer and their pitfalls: Is it possible to integrate the complexity of this neoplasm into a unified staging system?Critical Reviews in Oncology/Hematology, 2022
- Locoregional tumor burden and risk of mortality in metastatic breast cancernpj Precision Oncology, 2022
- Clinical information and management status of de novo stage IV breast cancer patients: a Chinese multicenter investigation (CSBrS-002)Chinese Medical Journal, 2021
- Metastatic relapse of stage I–III breast cancer in New ZealandCancer Causes & Control, 2021
- Breast cancer liver metastasis: current and future treatment approachesClinical & Experimental Metastasis, 2021
- Comparative evaluation of nuclear and histological grades as prognostic factors for invasive breast cancerBreast Cancer, 2020
- Behavior of metastatic breast cancer according to subtypeBreast Cancer Research and Treatment, 2020