Background: In patients with diffuse large B-cell lymphoma (DLBCL), central nervous system (CNS) relapse is uncommon but is nearly always fatal. This study aimed to determine the risk factors for CNS relapse in DLBCL patients and to evaluate the efficacy of rituximab and intrathecal chemotherapy prophylaxis for CNS relapse reduction. Methods: A total of 511 patients with newly diagnosed DLBCL treated at the Sun Yat-sen University Cancer Center between January 2003 and December 2012 were included in the study. Among these patients, 376 received R-CHOP regimen (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) as primary treatment, and 135 received CHOP regimen (cyclophosphamide, doxorubicin, vincristine, and prednisone) as primary treatment. Intrathecal chemotherapy prophylaxis (methotrexate plus cytarabine) was administered to those who were deemed at high risk for CNS relapse. In the entire cohort and in the R-CHOP set in particular, the Kaplan-Meier method coupled with the log-rank test was used for univariate analysis, and the Cox proportional hazards model was used for multivariate analysis. Differences were evaluated using a two-tailed test, and P < 0.05 was considered significant. Results: At a median follow-up of 46 months, 25 (4.9%) patients experienced CNS relapse. There was a trend of reduced occurrence of CNS relapse in patients treated with rituximab; the 3-year cumulative CNS relapse rates were 7.1% in CHOP group and 2.7% in R-CHOP group (P = 0.045). Intrathecal chemotherapy prophylaxis did not confer much benefit in terms of preventing CNS relapse. Bone involvement [hazard ratio (HR) = 4.21, 95% confidence interval (CI) 1.38-12.77], renal involvement (HR = 3.85, 95% CI 1.05-14.19), alkaline phosphatase (ALP) > 110 U/L (HR = 3.59, 95% CI 1.25-10.34), serum albumin (ALB) < 35 g/L (HR = 3.63, 95% CI 1.25-10.51), treatment with rituximab (HR = 0.34, 95% CI 0.12-0.96), and a time to complete remission = 108 days (HR = 0.22, 95% CI 0.06-0.78) were independent predictive factors for CNS relapse in the entire cohort. Bone involvement (HR = 4.44, 95% CI 1.08-18.35), bone marrow involvement (HR = 11.70, 95% CI 2.24-60.99), and renal involvement (HR = 10.83, 95% CI 2.27-51.65) were independent risk factors for CNS relapse in the R-CHOP set. Conclusions: In the present study, rituximab decreased the CNS relapse rate of DLBCL, whereas intrathecal chemotherapy prophylaxis alone was not sufficient for preventing CNS relapse. Serum levels of ALB and ALP, and the time to complete remission were new independent predictive factors for CNS relapse in the patients with DLBCL. In the patients received R-CHOP regimen, a trend of increased CNS relapse was found to be associated with extranodal lesions.