AbstractAbstract 2814 Background:Central nervous system (CNS) relapse is considered an infrequent but severe, nearly fatal complication of diffuse large B-cell lymphoma (DLBCL) following initial chemotherapy. Intrathecal (IT) prophylaxis cannot be recommended for all DLBCL patients because of the low probability of CNS relapse. Rituximab (R) added to CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) has been reported as likely to reduce the frequency of CNS relapse in patients with DLBCL, however, the efficacy of rituximab on the prognosis of CNS relapse compared with those who relapsed at other sites and predictive factors for CNS relapse in patients treated with rituximab contained chemotherapy remain unclear. The main objective was to determine the high-risk patients who need IT prophylaxis and the outcomes of CNS relapse compared with relapse on other sites. Patients and Methods:Diffuse large B cell lymphoma patients treated with rituximab contained CHOP regimen with or without radiotherapy and IT prophylaxis at Cancer Institute Hospital of JFCR between October 2003 and December 2006 were analyzed retrospectively. CNS relapse was defined as leptomeningeal, brain parenchymal, or intradural involvement with lymphoma, based on radiological findings or the presence of malignant lymphoma cells in spinal fluid. Results:A total number of 137 patients were identified. IT prophylaxis was administered in 13 of 137 patients (9.48%) depending on the sites of lymphoma involvement, such as bone marrow involvement, testis, paranasal sinuses. 17 of 137 patients (12.4%) underwent radiotherapy after chemotherapy because of early stage or their residual disease. With a median follow-up period of 4.4 years, 9 patients had experienced CNS relapse (6.7%: 3.0–12.1, 95%CI) among 30 documented relapses, with 9 presenting with nodal relapse alone and 21 presenting with extranodal relapse including CNS. IT prophylaxis and the addition of radiotherapy did not affect the frequency of CNS relapse (P=0.6 and 0.26). Median time to CNS relapse was 20 months. Overall survival (OS) was significantly inferior in CNS relapse patients to other-sites relapse patients, (median OS, 61 months vs. did not occur; P =.042). Nevertheless, OS was not significantly different between patients with CNS relapse or at other sites. In univariate analysis, factors associated with CNS relapse (P < 0.05) included age over 65 years and serum levels of soluble IL-2 receptors (sIL-2R) ≧10 ULN (upper limit of normal) but not sex, PS ≧3, stage ≧4, B symptom, bulky mass, elevated LDH >2 ULN, elevated MIB1 index ≧90%, poor revised-international prognostic index (R-IPI), extranodal sites ≧2, or type of GC or non-GC. Multivariate Cox regression analysis identified increased serum levels of (sIL-2R) (P =0.037) as an independent predictive factor for CNS relapse (P=0.04, HR=7.02: 95%CI=1.87-26.22). Five of 9 CNS relapse patients were still alive with the combination treatment of whole brain irradiation, systemic chemotherapy (R- dexamethasone, cisplatin, cytarabine) and intrathecal chemotherapy. Conclusions:The incidence rate of CNS relapse in 137 DLBCL patients treated with R-CHOP, CEOP regimens may be lower than with CHOP in agreement with previous studies. Furthermore, rituximab may improve OS after CNS relapse. Ten times increased serum s-IL2R is a potential independent risk factor for CNS relapse and should be included in the IT prophylaxis indication in patients with DLBCL. Disclosures:No relevant conflicts of interest to declare.