Multi-omics liquid biopsy for leptomeningeal metastases in lung cancer: from molecular mechanisms to precision diagnosis and treatment.

Lung cancer remains the leading cause of cancer-related deaths worldwide. Leptomeningeal metastasis (LM), a serious complication of advanced lung cancer, poses significant challenges in both clinical diagnosis and treatment monitoring, and is associated with a poor prognosis. Conventional diagnostic methods include magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) cytology; however, their sensitivity is limited, particularly in the early stages of the disease. Against this backdrop, liquid biopsy has emerged as a minimally invasive and practical approach for molecular assessment and longitudinal monitoring of LM. Recent studies have reported that, in selected cohorts of patients with non-small cell lung cancer (NSCLC) and LM, CSF-derived circulating tumor DNA (ctDNA) detects driver and resistance-associated alterations more frequently than paired plasma, while epithelial cell adhesion molecule (EpCAM)-based CSF circulating tumor cell (CTC) assays have reported sensitivities of 76%-100% and specificities of 84%-100%, compared with a first-lumbar-puncture sensitivity of 44%-67% for conventional CSF cytology. These performance estimates should be interpreted in the context of selected cohorts and heterogeneous reference frameworks. Furthermore, cerebrospinal fluid circulating tumor cell (CSF-CTC) burden holds prognostic value, with higher CTC levels associated with poorer survival outcomes. This review summarizes recent advances in liquid biopsy biomarkers including ctDNA, DNA methylation, microRNA (miRNA), proteins, and circulating tumor cells (CTCs), focusing on their applications in the diagnosis, treatment monitoring, and prognostic assessment of lung cancer-associated LM, and provides a comparative discussion of liquid biopsy based on plasma and cerebrospinal fluid specimens. It further explores the potential, current challenges, and directions for clinical translation of multi-omics integration in the personalized management of LM.