Some patients with epithelial-cell cancers develop leptomeningeal carcinomatosis (LC), a severe complication difficult to diagnose and with an adverse prognosis. This study explores the contribution of flow cytometry immunophenotyping (FCI) to the diagnosis and prognosis of LC. Cerebrospinal fluid (CSF) samples from patients diagnosed with LC were studied using FCI. Expression of the epithelial-cell adhesion molecule (EpCAM) was the criterion used to identify the epithelial cells. To test the diagnostic precision, 144 patients (94 diagnosed with LC) were included. The prognostic value of FCI was evaluated in 72 patients diagnosed with LC and eligible for therapy. Compared with cytology, FCI showed greater sensitivity and negative predictive value (79.79 vs. 50%; 68.85 vs. 51.55%, respectively), but lower specificity and positive predictive value (84 vs. 100%; 90.36 vs. 100%, respectively). The multivariate analysis revealed that the percentage of CSF EpCAM+ cells predicted an increased risk of death (HR: 1.012, 95% CI 1.000-1.023; p=0.041). A cut-off value of 8% EpCAM+ cells in the CSF distinguished two groups of patients with statistically significant differences in overall survival (OS) (p=0.018). This cut-off value kept its statistical significance regardless of the absolute CSF cell-count. The FCI study of the CSF improved the sensitivity for diagnosing LC, but refinement of the technique is needed to improve specificity. Furthermore, quantification of CSF EpCAM+ cells was revealed to be an independent prognostic factor for OS in patients with LC eligible for therapy. An 8% cut-off value contributed to predicting clinical evolution before initiation of therapy.