Abstract PO5-03-03: Nomogram for predicting brain metastasis in early-stage breast cancer patients

Abstract Nomogram for predicting brain metastasis in early-stage breast cancer patients Background: Brain metastasis is a significant concern for patients with early-stage breast cancer, as it can have a detrimental impact on survival and quality of life. Clinical and pathological characteristics of the primary tumor may be modeled into a nomogram that can predict the risk of brain metastasis. The goal is to create a nomogram and clinical calculator to assist in selection of patients with high risk of brain metastasis to enrich patient population for accrual on clinical trials for the purpose of preventing or delaying emergence of brain metastasis, at an early stage of their disease. Methods: We reviewed all early-stage breast cancer patients (stages I to III) N=40290 from the University of Texas MD Anderson Cancer Center, Houston, TX, from 1/1/1997 to 5/8/2020. We utilized multiple predictive statistical models to identify the most relevant variables and optimal predictive criteria. Variables for the model included patient age, tumor characteristics (such as grade,hormone receptor and HER2 status) and medical treatment (such as chemotherapy and endocrine therapy). The Receiver Operating Characteristic (ROC), and its AUC (area under the curve) was used to demonstrate the performance of the multivariate model to predict the occurrence of brain metastasis at different time points (1 year, 2 years, and 5 years), distinguishing between patients who develop brain metastasis in a time-dependent manner and those who do not. Results: The results showed the following clinical variables to be statistically significant in predicting brain metastasis: younger age, high estrogen and progesterone receptor percentage, higher tumor size; use of aromatase inhibitor therapy, and use of selective estrogen receptor modulator therapy were associated with a decreased risk of brain metastasis. Higher grade, Ki-67 levels, HER2-positive status, lymphovascular invasion and the use of chemotherapy was associated with an increased risk. The AUC values for the prediction of brain metastasis at different time points are as follows: 0.85 at 1 year, 0.83 at 2 years, and 0.82 at 5 years. These values indicate that the multivariate model has good discriminative ability, with higher values indicating a better predictive performance. An AUC of 0.85 suggests that the model has a high probability of correctly distinguishing between patients who will develop brain metastasis and those who will not at 1 year. Conclusions: Our predictive tool holds promise in assisting clinicians when making informed, personalized care decisions for patients at risk of developing brain metastasis. The model will be validated using independent patient cohorts from Institut Jules Bordet, Brussel, Belgium, to assess their accuracy and reproducibility. Continued research and validation will further refine these models, increasing their reliability and their clinical/clinical research applicability. Citation Format: Akshara Singareeka Raghavendra, Nuria Kotecki, Kristofer Jennings, Ottavia Amato, Debu Tripathy, Ahmad Awada, Nuhad Ibrahim. Nomogram for predicting brain metastasis in early-stage breast cancer patients abstract. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-03-03.