Brain metastases (BM) are the most common intracranial tumours in adults and a major source of morbidity and mortality. While outcomes have historically been poor, a marked disparity exists in the propensity for different primary cancers to spread to the brain (organotropism). To systematically quantify this risk, we analysed nearly 3.8 million primary cancer cases and almost 56,000 cases of de novo (synchronous) BM from the population-based Surveillance, Epidemiology, and End Results (SEER) registry.
Our analysis found that the clinical burden is highly concentrated, with five primary cancers—lung, breast, melanoma, kidney, and colorectal—accounting for over 90% of all de novo BM cases. Multivariable logistic regression and machine learning models were used to identify key clinical predictors, revealing that lymph node involvement (regional or distant) and histology were generally more informative than tumour grade or T stage. Nodal spread was specifically associated with increased brain tropism across cancer types. Even after adjusting for clinical variables, tumour-intrinsic biology remained dominant: lung cancer retained a ~47-fold higher likelihood of presenting with BM than prostate cancer, indicating that biological programmes specific to certain primaries drive neurotropism.
Analysis of high-risk subgroups revealed BM prevalence exceeding 10% at diagnosis among select populations, including HER2-positive and triple-negative breast cancer, advanced stage lung cancer, kidney cancers with multiple metastases, and metastatic melanoma. Risk was further amplified when extra-cranial metastases and distant nodes were present. These data support targeted, risk-adapted surveillance for BM rather than universal screening.
To unlock the molecular drivers underpinning this biological preference, we initiated a multi-omic investigation. We have banked >125 fresh-frozen BM with matched germline and completed WGS, RNA-seq and methylation profiling on 60 cases. Early analyses suggest that molecular drivers vary by primary tumour, and that rarer BM show alterations in pathways likely to influence migration, invasion and adaptation to the brain microenvironment.
These findings provide a framework for integrating epidemiological data, personalized risk modeling, and mechanistic cancer biology to improve BM prediction and inform clinical surveillance protocols. Future research should focus on validating predictive models in prospective cohorts and investigating pathobiological mechanisms driving brain-tropic metastasis in solid tumors.