RADT-87. PREDICTORS OF LEPTOMENINGEAL METASTASIS FOLLOWING BRAIN PARENCHYMAL METASTASES IN NSCLC PATIENTS: A COMPETING RISKS ANALYSIS

Abstract BACKGROUND Leptomeningeal metastasis (LM) is a fatal complication in non-small cell lung cancer (NSCLC), with increasing incidence due to improved treatments and survival. While LM typically develops after brain metastases (BM), its predictors largely remain unknown. This study aims to identify predictors of LM in NSCLC patients with BM using a competing risks approach. METHODS We analyzed 112 NSCLC patients with BM treated at Sanjiu Brain Hospital (2014-2020) without prior whole brain radiation therapy. MR images were independently reviewed by two radiologists. The proximity of BM lesions to the subdural space/ventricles was measured and stratified into invading/neighboring or distant lesions. Competing risks analysis was performed using R software (version 4.1.2) with Gray’s test comparing incidence between groups and Fine-Gray regression identifying independent predictors. RESULTS The cohort (42.8% male, 57.2% female; median age 54 years; median follow-up 9.9 months) showed significantly higher LM incidence in mutation-positive patients (63.5% vs 27.6% at 1500 days, p=0.006). Proximity to CSF strongly predicted LM development, with patients having metastases more proximal to CSF spaces showing significantly higher cumulative incidence (67.9% [95% CI: 54.8-81.1%] vs 18.4% [95% CI: 7.5-29.6%], p<0.001). SRS treatment showed protective effects (38% vs 92.8% LM incidence, p<0.001). Fine-Gray regression confirmed mutation status (HR=2.988, 95% CI: 1.03-8.705, p=0.045), proximity to CSF (HR=3.166, 95% CI: 1.68-5.966, p<0.001), and SRS treatment (HR=0.255, 95% CI: 0.14-0.465, p<0.001) as independent predictors. CONCLUSION This competing risks analysis suggests NSCLC patients with BMs with mutation-positive status and/or CSF proximal BM lesions should receive intensive surveillance, and SRS can be considered as a preventive strategy against LM spread. Our competing risks approach provides more accurate risk estimates than conventional survival analysis methods. However, future prospective studies are warranted to validate these findings and develop risk-stratified protocols for patients with brain metastases at risk for leptomeningeal spread.