Cerebrospinal fluid biomarker profiling in patients with non–small cell lung cancer and leptomeningeal disease.

2026 Background: Cerebrospinal fluid (CSF) cytology remains the standard for diagnosing leptomeningeal disease (LMD) and monitoring treatment response, but its clinical utility is limited by low sensitivity and nonquantitative results. The role of novel CSF biomarkers, such as circulating tumor cells (CTCs), in improving disease assessment in non–small cell lung cancer (NSCLC) with LMD is not fully defined. Methods: We performed retrospective clinical and genomic analysis of patients with NSCLC and LMD who underwent CSF CTC testing using the CellSearch platform at Memorial Sloan Kettering between 1/1/2015 and 11/5/2025. Time-to-event analyses were calculated from initial positive CSF CTC measurement (≥3 cells/3mL) or from completion of proton craniospinal irradiation (pCSI). Optimal CTC cutoffs were determined using recursive partitioning analysis (RPA). CSF cytologic clearance was defined as an initial positive cytology followed by two consecutive negative results at least 4 weeks apart. Next-generation sequencing was performed using MSK-IMPACT. Results: Among 148 patients with NSCLC and LMD, 72% (n=105) were female, 53% (n=78) were never-smokers, and 92% (n=136) harbored an oncogenic driver alteration, most commonly epithelial growth factor receptor ( EGFR) mutations (60%, n=89). Fourteen percent (n=20) were diagnosed with LMD at time of metastatic disease, whereas 54% (n=80) had received 2 or more prior lines of therapy. Sixty-two percent (n=92) received radiation therapy for LMD, including 43% (n=64) undergoing pCSI and 18% (n=26) undergoing photon radiation (involved field radiation therapy [IFRT] to brain/spine, n=10; whole brain radiation therapy [WBRT] alone, n=10; WBRT + IFRT to spine, n=4; photon CSI, n=2). With a median follow-up of 21.6 months (95% CI 16.2–37.9), patients with a lower baseline CSF CTC count (≤90 CTCs/3mL, n=66; cutoff identified through RPA) had longer median overall survival (mOS; 10.2 vs 6.6 months; HR 0.53, 95% CI 0.35–0.79; p=0.002) compared with those with higher counts (>90 CTCs/3mL, n=82). Among patients who underwent pCSI with serial CSF testing (n=34), a ≥30% reduction in CTCs was associated with a longer mOS (13.1 vs 6.2 months; HR 0.15, 95% CI 0.05–0.47; p=0.0001) compared with those with <30% reduction; only 12% (n=4) achieved CSF cytologic clearance. In EGFR- mutant NSCLC, genomic analysis of CSF samples with higher baseline CTC count (>90 CTCs/3mL, n=24) demonstrated a higher frequency of TP53 (p=1.559e-3) and MTAP (p=8.31e-3) alterations, and a lower frequency of MDM2 (p=0.05) alterations, compared to patients with lower baseline CTC counts (≤90 CTCs/3ml, n=23). Conclusions: Lower baseline CSF CTC counts and reductions following pCSI are associated with improved survival in patients with NSCLC and LMD. CSF CTC assessment provides a quantitative measure of LMD burden and may complement conventional CSF cytology.