Development and Validation of a Novel Conditional Survival Nomogram for Predicting Real-Time Prognosis in Patients With Breast Cancer Brain Metastasis.

Abstract

Background: Breast cancer brain metastasis (BCBM) prognosis has not been evaluated dynamically, which may underestimate patient survival. This study aimed to perform a conditional survival (CS) analysis and develop and validate an individualized real-time prognostic monitoring model for survivors.

Methods: The study included patients with BCBM from the Surveillance, Epidemiology, and End Results database (training group, n = 998) and our institution (validation group, n = 45) and updated patient overall survival (OS) over time using the CS method: CS(t2|t1)=OS(t1+t2)OS(t1). Multivariate Cox regression was used to identify prognostic factors for the nomogram, which estimated individualized OS. Furthermore, a novel CS-nomogram and its web version were further developed based on the CS formula.

Results: CS analysis showed that the 5-year OS of BCBM survivors gradually improved from 13.5% estimated at diagnosis to 26.0%, 39.7%, 57.9%, and 77.6% (surviving 1-4 years, respectively). Cox regression identified age, marital status, estrogen receptor status, human epidermal growth factor receptor 2 (Her-2) status, histological grade, surgery, and chemotherapy as significant factors influencing OS (P < .05). We then constructed and deployed the CS-nomogram based on the CS formula and the nomogram to predict real-time prognosis dynamically (https://wh-wang.shinyapps.io/BCBM/). During performance evaluation, the model performed well in both the training and validation groups.

Conclusions: CS analysis showed a gradual improvement in prognosis over time for BCBM survivors. We developed and deployed on the web a novel real-time dynamic prognostic monitoring system, the CS-nomogram, which provided valuable survival data for clinical decision-making, patient counseling, and optimal allocation of healthcare resources.