Prediction of brain metastases of breast cancer using next generation sequencing genomics.

e14004 Background: Breast cancer was the most common malignancy diagnosed in 2023 with over 300,000 new diagnoses. Though it is highly treatable when caught in early stages, metastatic disease is highly morbid. Approximately 0.4% of all cases and 7.2% of cases of metastatic breast cancer will develop brain metastases. Hormonal receptors correlate poorly with risk of brain metastasis. Predictive factors for breast cancer brain metastases remain elusive. This project aimed to identify genomic risk factors for the metastasis of breast cancer to the brain. Methods: We created a retrospective cohort of all breast cancer patients with next generation sequencing (NGS) at a comprehensive cancer center and its associated clinics. All NGS in the cohort was performed by FoundationOne. Brain metastases were defined via ICD-10 codes. Univariate analysis was performed using Fisher’s Exact test. Significance was defined as p<0.05 after adjustment of multiple testing corrections performed using the Benjamini-Hockberg procedure. A multivariable logistic regression model was fit; to avoid overfitting, the maximum number of covariables was defined by requiring 10 cases for every variable included. Model fit was compared by the Akaike information criterion (AIC) statistic. SAS v9.4 was used for all analyses. Results: This final cohort included 242 patients, of which 29 (12%) developed brain metastases. Among the standard panel of 223 tested genes, 11 were significantly associated with brain metastases after multiple testing correction. Significant genes included: TP53, GATA3, PIK3CA, EP300, MSH3, NTRK1, BRCA1, CBFB, FGFR4, MAP2K2, and NOTCH2. TP53 was the most prevalent gene mutation amongst brain metastases patients. The gene was present in 79% (23/29) of all brain metastases cases and 96% (23/24) of patients with TP53 had brain metastasis. PIK3CA and GATA3 were the next most common in 10 (34%) and 6 (21%) cases respectively. In building predictive models, the optimal model included TP53, GATA3, and CBFB which presented an AIC of 33.15 and an AUC of 0.98. The best two variable model included TP53 and PIK3CA with AIC of 49.67 and AUC of 0.95. Conclusions: Among this cohort, 11 genes were found to be significantly associated with the development of brain metastases. Among them, TP53 was the most prevalent while GATA3, PIK3CA, and CBFB also contributed to predictive ability. These variables suggest high levels of predictability, though given the limited sample thorough validation will be required. However, given the numerous genes found to be associated and the significant starting predictivity, there is potential for NGS to identify those at higher risk of brain metastases. Further study and expansion of NGS will be needed to validate these results and investigate the timing of brain metastases for the potential of early detection.