Is there still a need for specific central nervous system directed prophylaxis for diffuse large B-cell lymphoma in the rituximab era?

Abstract

Central nervous system (CNS) relapse remains an important cause of treatment failure in patients with diff use large B-cell lymphoma (DLBCL) treated with rituximab, cyclophosphamide, doxorubicin, vincristine and prednisolone (R-CHOP). Th e frequency with which it occurs depends upon the population studied, the duration of follow-up and the degree of scrutiny applied. In unselected patients with DLBCL, CNS relapse rates of 2 – 9.9% have been reported [1,2], whilst in a selected high-risk minority the actuarial 2-year incidence is as high as 33.5% [3]. Although infrequent, CNS relapse has devastating consequences and is rarely salvageable; hence, prevention is the best approach. However, the use of, appropriate selection of patients for and most eff ective form of prophylaxis all remain contentious issues. Rituximab (Mabthera; Hoff man La-Roche, Basel, Switzerland; and Rituxan; Genentech, San Francisco, CA) has markedly improved outcomes in patients with DLBCL [4,5], but prior studies evaluating its impact on CNS relapse have had confl icting results. Helping to clarify our understanding, in this issue of Leukemia and Lymphoma Zhang et al . report a metaanalysis of eight randomized studies comprising 4911 patients, in which patients with newly diagnosed DLBCL received chemotherapy with or without rituximab [6]. Th e risk factors for CNS relapse identifi ed in this study included a list of the usual suspects: advanced stage, International Prognostic Index (IPI) 2, performance status 1, elevated serum level of lactate dehydrogenase (LDH), more than one extranodal site of involvement, bone marrow involvement, the presence of B symptoms and testicular involvement. Intrathecal (IT) prophylaxis was not associated with reduced risk of CNS relapse, although in the included studies use of IT chemotherapy was not randomly allocated, and was presumably used in those patients perceived to be at higher risk of CNS relapse; thus the apparent equivalent risk for CNS relapse in patients receiving IT prophylaxis (hazard ratio [HR] 1.03) could be interpreted as consistent with some degree of risk reduction, although even if present such a benefi t is insuffi cient on its own. [7] Th e main fi nding of the study was that the addition of rituximab showed a modest reduction of CNS relapse in DLBCL, from 5.7% to 4.7%, a relative risk reduction of 30% (HR 0.7; 95% confi dence interval [CI] 0.54 – 0.91), or an absolute risk reduction of 1% in the total population. How should we incorporate the fi ndings from this into clinical practice? It appears that the addition of rituximab to CHOP aff ords a modest reduction in CNS relapse, and this likely provides suffi cient protection for those patients without identifi ed high risk factors. However, we argue that for patients at high risk of CNS relapse (predicted 10%), the use of R-CHOP without specifi c CNS-directed prophylaxis is insuffi cient. A number of the risk factors identifi ed in this meta-analysis are markers of high tumor burden and/or highly proliferative disease. In addition to the risk factor of testicular involvement identifi ed in this metaanalysis, other retrospective studies have identifi ed specifi c high-risk extranodal sites including kidney [8], breast [9,10], adrenal gland [10] and epidural space [11,12]. We believe that patients with one or more of these factors also warrant consideration of CNS-directed therapy. Although paranasal sinus involvement increased risk of CNS dissemination in the pre-rituximab era, a recent German analysis of patients treated with R-CHOP has found no such increase [13]. Th is meta-analysis confi rms the fi nding of prior studies that IT chemotherapy alone as prophylaxis is either ineff ective or has insuffi cient utility [3,7 – 9,14]. Th e practice of IT prophylaxis is extrapolated from its successful use in acute lymphoblastic leukemia and Burkitt lymphoma, where the risk of CNS involvement is high and the distribution is leptomeningeal in 74% of instances [15]. In contrast, meta-analysis of 28 studies of DLBCL found that CNS relapse was manifest as parenchymal involvement in 59% of cases [16]. Th is, combined with uneven distribution of chemotherapy injected by lumbar puncture around the