Genomic Predictors of Leptomeningeal Disease Development among Patients with Brain Metastases

Abstract

Leptomeningeal disease (LMD) is associated with significant neurologic symptomatology, functional decline, and generally, a very poor prognosis. Clinical characteristics of patients with parenchymal brain metastases have limited potential in predicting who will subsequently develop LMD. We hypothesized that genomic alterations may predict which patients with intracranial disease are at highest risk for developing LMD and sought to identify DNA-based genomic alterations among a targeted panel of cancer-related genes that may increase a patient's risk for LMD. We identified 810 patients with parenchymal brain metastases secondary to solid tumor primaries without LMD at diagnosis of initial intracranial disease managed at a tertiary cancer center (2003-2020) for whom next-generation sequencing panel data (OncoPanel, 239 genes) were available on at least one extracranial or intracranial tumor specimen. Fine/Gray's competing risks regression was utilized to compare risk for LMD development among patients with vs. without somatic alterations of likely clinical/biological significance, delineated via OncoKB, across 96 genes with a mutational frequency ≥0.5% in the patient cohort. Genes with a q-value<0.10 and hazard ratio (HR)>1 were considered predictive of LMD risk; patients were dichotomized into "high-risk" vs. "low-risk" of LMD development based on the presence or absence of mutations in any one of these predictive genes. Genomic alterations of potential biological significance in MAPK1 (gain-of-function), CDH1 (loss-of-function), and SF3B1 (switch-of-function) were more common among patients who developed LMD vs. not (MAPK1, 3.6% vs. 0.4%; CDH1, 3.6% vs. 0.8%; SF3B1, 3.6% vs. 0.6%, respectively) and were each associated with an increased risk for LMD development (q-value<0.10 in all cases). On multivariable Fine/Gray's competing risk regression, "high-risk" patients with genomic alterations in any of these three genes (HR 4.32 [1.93-9.67], p<0.001), Karnofsky performance status <90 (HR 1.76 [1.11-2.79], p<0.001), and lack of local, brain-directed therapy as part of intracranial disease management (HR 2.97 [1.48-5.96], p<0.002) were associated with increased risk of LMD; age and primary tumor site were not associated with LMD risk (p>0.05). Utilizing a targeted panel of genes with a known role in cancer pathogenesis, we identified genomic alterations in three genes as being predictive of LMD development. If validated in independent datasets, development of clinical trials exploring inhibition of pathways affected by these genomic alterations may be warranted with the goal of LMD prevention and targeted treatment among particularly high-risk cohorts. To our knowledge, this represents the first study to identify potentially actionable alterations as predictive of leptomeningeal disease development among patients with brain metastases.