Purpose: Extranodal natural killer (NK)/T cell lymphoma, nasal type frequently involves paranasal sinuses. Although involvement of the paranasal sinuses has been regarded as a possible risk factor for central nervous system (CNS) relapse, its incidence was variable from 0% to 6% in extranodal NK/T cell lymphoma. Thus, an exact incidence of CNS relapse and a definitive role of CNS prophylaxis are still controversial in extranodal NK/T cell lymphoma, nasal type. Therefore, we performed this study to assess the incidence of CNS relapse and identify risk factors that predict CNS relapse to establish guidelines for CNS prophylaxis in patients with extranodal NK/T cell lymphoma.Patients and methods: We analyzed 208 patients with extranodal NK/T cell lymphoma, nasal type from June 1995 to March 2008. CNS relapse was defined as a brain parenchyma lesion or leptomeningeal relapse during the follow-up after a complete response (CR) or CNS involvement during the first-line therapy or salvage chemotherapy. For subgroup analysis, we used the international prognostic index (IPI), and NK/T cell prognostic index (NKPI), proposed previously for extranodal NK/T cell lymphoma, which includes the presence of B symptoms, stage III or IV, elevated serum LDH concentration, and lymph node involvement. We designated group III/IV patients (those in group III with 2 risk factors and those in group IV with 3 or 4 risk factors) as a high risk group. Second, we dichotomized patients into two groups based on the primary site of involvement. Upper aerodigestive tract NK/T cell lymphoma (UNKTL) included all lymphomas confined to the nasal cavity, nasopharynx, and upper aerodigestive tract including the oral cavity, larynx, and pharynx. NK/T cell lymphomas involving all other sites were defined as extra-upper aerodigestive tract NK/T cell lymphoma (EUNKTL).Results: Of the 208 patients, 12 (5.76%) patients experienced CNS relapse (median 11.62 months, range 0.2–123.2 months). The median time to CNS relapse was 6.03 months (95% confidence interval: 5.23–6.83 months). Eight patients showed CNS relapse during or immediately after the completion of the first-line treatment while four patients showed CNS relapse combined with systemic relapse. The pattern of CNS relapse was balanced: 6 had lesions of brain parenchyma and 6 leptomeningeal relapse. The prognosis after CNS relapse was poor even though some patients showed complete resolution of the CNS lesion because systemic disease progression followed in most patients. Four responders to CNS-directed therapy showed disease progression, and one responder was lost during follow-up. Even though all patients received CNS-directed therapy including intrathecal chemotherapy or whole-brain radiotherapy combined with systemic chemotherapy, most patients died within 6 months after CNS relapse because of disease progression. Age, sex, Eastern Cooperative Oncology Group (ECOG) performance, the presence of B symptoms, serum LDH concentration, bone marrow invasion, the primary site of involvement, and IPI risk were balanced between patients with and without CNS relapse (p > 0.05). The clinical variables associated with CNS relapse were as follows: Ann Arbor stage III or IV (15.87%, P < 0.001), presence of regional lymph node involvement (10.41%, P = 0.006), group III/IV of NKPI (NK/T cell lymphoma prognostic index, 10.20%, P = 0.003), and extra-upper aero-digestive primary sites (9.75%, P = 0.008). In multivariate analysis, NKPI retained the strongest statistical power to predict CNS relapse (p = 0.007, relative risk 9.289, 95% confidence interval: 1.828–47.212) in extranodal NK/T cell lymphoma, nasal type. Thus, the risk of CNS relapse might be higher in patients with group III/IV of NKPI than group I/II of NKPI.Conclusion: Despite NK/T cell lymphoma frequently involves paranasal sinus, a routine CNS evaluation and prophylaxis do not seem to be necessary in NKPI group I or II patients due to a very low incidence. However, CNS prophylaxis should be considered in NKPI groups III and IV.