BACKGROUND Leptomeningeal metastases (LM) are increasingly diagnosed as anticancer therapies become more effective and result in prolonged patient survival. OBJECTIVE To evaluate survival, cause of death, and treatment-related toxic effects in patients undergoing combined modality therapy for LM of non-small cell lung cancer. METHODS Thirty-two patients (age range, 48-73 years; median, 57 years) with LM attributable to metastatic non-small cell lung cancer were treated prospectively. Neurologic presentation included headache (11 patients), cranial neuropathies (9), ataxia (5), cauda equina syndrome (3), myelopathy (3), meningismus (2), radiculopathy (2), and confusion (1). All patients underwent radiographic evaluation to determine the extent of central nervous system disease followed by radiotherapy (16 patients) and sequential and intraventricular chemotherapy (methotrexate in 32 patients; cytarabine in 16; and thiotepa in 6). Twelve patients received concurrent systemic chemotherapy. RESULTS Central nervous system imaging demonstrated interrupted cerebrospinal fluid flow (13 patients), parenchymal brain metastases (9), subarachnoid nodules (8), hydrocephalus (5), and epidural spinal cord compression (2). Cytological responses were seen in 17 patients to first-line chemotherapy, 8 to second-line chemotherapy, and 2 to third-line chemotherapy. Treatment-related toxic effects included 20 patients with aseptic meningitis (grade 2 in 16; grade 3 in 4) and 12 patients with grade 3 or 5 thrombocytopenia or neutropenia (4 related to intraventricular chemotherapy). Median survival was 5 months (range, 1-12 months). Nineteen patients died of progressive LM or combined LM and systemic disease progression. Patients with persistent interruption of cerebrospinal fluid flow fared worse than patients with normal cerebrospinal fluid flow (median survival, 4 vs 6 months; P<.05). CONCLUSIONS Leptomeningeal metastases in patients with non-small cell lung cancer may be palliated with combined modality therapy; however, therapy and survival is based on the extent of central nervous system disease present at pretreatment evaluation.