Leptomeningeal carcinomatosis (LC) is a metastatic cancer invading the central nervous system (CNS). We previously reported a metabolomic diagnostic approach as tested on an animal model and compared with current modalities. Here, we provide a proof of concept by applying it to human LC originating from lung cancer, the most common cause of CNS metastasis. Cerebrospinal fluid from LC (n 5 26) and normal groups (n 5 41) were obtained, and the diagnosis was established with clinical signs, cytology, MRI and biochemical tests. The cytology on the CSF, the current gold standard, exhibited 69% sensitivity (~100% specificity) from the first round of CSF tapping. In comparison, the nuclear magnetic resonance spectra on the CSF showed a clear difference in the metabolic profile between the LC and normal groups. Multivariate analysis and crossvalidation yielded the diagnostic sensitivity of 92%, the specificity of 96% and the area under the curve (AUC) of 0.991. Further spectral and statistical analysis identified myo-inositol (p < 5 3 10 214 ), creatine (p < 7 3 10 28 ), lactate (p < 9 3 10 24 ), alanine (p < 7.9 3 10 23 ) and citrate (p < 3 3 10 24 ) as the most contributory metabolites, whose combination exhibited an receiver-operating characteristic diagnostic AUC of 0.996. In addition, the metabolic profile could be correlated with the grading of radiological leptomeningeal enhancement (R 2 5 0.3881 and p 5 6.66 3 10
24), suggesting its potential utility in grading LC. Overall, we propose that the metabolomic approach might augment current diagnostic modalities for LC, the accurate diagnosis of which remains a challenge.Leptomeningeal carcinomatosis (LC) is a devastating complication of systemic cancer that occurs in 5-10% of patients with solid tumors, more frequently adenocarcinoma, and is observed most commonly in patients with breast cancer, lung cancer or melanoma.1 Especially, lung cancer is one of the most common primary cancers that cause central nervous system (CNS) metastases, and the incidence of LC is increasing as lung cancer patients survive longer with more advanced treatment modalities. [2][3][4][5] In addition, with the expansion of therapeutic options for systemic treatment of lung cancer, including targeted therapies, the landscape for treating LC is changing. Although some drugs have demonstrated activity within the CNS, they have limitation to pass the blood-brain barrier. 6,7 Early and accurate diagnosis of LC and treatment initiation could offer the best chance of controlling symptoms and prevent the establishment of irreversible neurologic deficits that impair the patient's quality of life. 1,8,9 However, LC diagnosis currently remains a challenge. CSF cytology shows higher specificity (theoretical 100%) than MRI analysis (about 77%) and a sensitivity of 75% upon repeated spinal tapping.10,11 One