Wartość diagnostyczna i prognostyczna biomarkerów płynu mózgowo-rdzeniowego w przerzutach raka płuc do mózgu: analiza proteomiczna

Abstract Background Lung cancer brain metastasis (LCBM) is a severe complication with a poor prognosis, and reliable cerebrospinal fluid (CSF) biomarkers for its diagnosis are lacking. We aimed to discover and validate CSF protein markers that may aid the differential diagnosis of LCBM against common neurological conditions and to explore their diagnostic and prognostic value. Methods In a discovery cohort (n = 56) of LCBM patients and non-tumor controls undergoing lumbar puncture, high-precision proteomic mass spectrometry with permutation-based false discovery rate control screened differentially expressed proteins (DEPs). Candidate DEPs were validated by ELISA in a validation cohort (n = 78) comprising LCBM patients and disease controls with primary intracranial tumors, CNS lymphoma, encephalitis, cerebral hemorrhage, or epilepsy. Diagnostic performance was assessed by receiver operating characteristic analysis, with bootstrap validation and calibration of combined models. Prognosis was evaluated by Kaplan–Meier analysis (log-rank p adjusted for data-driven cutoff selection), Cox regression, and formal comparison of discriminative ability (C-index, time-dependent AUC, likelihood-ratio tests). Results ELISA confirmed differential expression of the candidates. KLK6 showed near-perfect discrimination of LCBM from all disease-control subgroups (AUC = 0.969), whereas sFRP4 and FBLN1 showed moderate accuracy. A combined model incorporating KLK6 achieved an AUC of 0.989 (95% CI, 0.976–1.000), and sFRP4 + FBLN1 yielded 0.797 (95% CI, 0.697–0.897; bootstrap-corrected 0.791). Higher CSF POSTN was associated with shorter overall survival (median, 19.0 vs. 51.0 months; HR = 2.05 per SD, 95% CI 1.25–3.37, p = 0.005), which persisted after adjustment for age (HR = 1.74 per SD, 95% CI 1.10–2.77, p = 0.019) and targeted therapy (HR = 0.56, 0.25–1.26, p = 0.163). Discrimination was higher for CSF POSTN than for CSF CEA (C-index, 0.749 vs. 0.577; difference, 0.171; 95% CI 0.016–0.303), although 36-month time-dependent AUC did not differ significantly; no superiority over CEA is claimed. Conclusion Integrating proteomics and immunoassays identified candidate CSF biomarkers differentiating LCBM from a spectrum of non-malignant and malignant neurological diseases. Because no NSCLC control group without brain metastasis was included, LCBM specificity has not been established; these markers should be regarded as preliminary auxiliary tools requiring independent external validation.