Central nervous system (CNS) events, including CNS relapse and progression to CNS, are known to be serious complications in the clinical course of patients with lymphoma. This study aimed to evaluate the risk of CNS events in patients with diffuse large B-cell lymphoma in the rituximab era. We performed a retrospective survey of Japanese patients diagnosed with diffuse large B-cell lymphoma who underwent primary therapy with R-CHOP chemoimmunotherapy between September 2003 and December 2006. Patients who had received any prophylactic CNS treatment were excluded. Clinical data from 1221 patients were collected from 47 institutions. The median age of patients was 64 years (range, 15–91 years). We noted 82 CNS events (6.7%) and the cumulative 5-year probability of CNS events was 8.4%. Patients with a CNS event demonstrated significantly worse overall survival (P < 0.001). The 2-year overall survival rate after a CNS event was 27.1%. In a multivariate analysis, involvement of breast (relative risk [RR] 10.5), adrenal gland (RR 4.6) and bone (RR 2.0) were identified as independent risk factors for CNS events. We conclude that patients with these risk factors, in addition to patients with testicular involvement in whom CNS prophylaxis has been already justified, are at high risk for CNS events in the rituximab era. The efficacy and manner of CNS prophylaxis in patients for each involvement site should be evaluated further. (Cancer Sci 2012; 103: 245–251)
The central nervous system (CNS) is thought to be a sanctuary for lymphoma cells from systemic chemoimmunotherapy, such as rituximab (R) plus CHOP (cyclophosphamide [CPA], doxorubicin [adriamycin, ADR], vincristine [VCR] and prednisolone [PSL]), because standard doses of these drugs do not adequately penetrate the CNS. Occurrence of a CNS event, defined as CNS relapse during systemic complete remission or CNS progression during concurrent systemic active lymphoma, is associated with extremely poor prognosis, with median survival of <6 months.(1–6) Many studies concerning CNS prophylaxis have been conducted; however, the efficacy of such prophylaxis in preventing CNS events is controversial.(5,7–12) The discrepancies between reports might be due to the differences in the various subtypes of lymphoma histology and the variability of treatment of CNS prophylaxis.(13–16) In addition, R has had a substantial impact on outcomes in patients with diffuse large B-cell lymphoma (DLBCL).(17) It is thus necessary to re-evaluate the risk of CNS events in the R era.
The present study comprises a multicenter retrospective analysis of patients with uniform DLBCL histology who have undergone uniform treatment with R-CHOP, widely accepted as the standard therapy in the R era. Patients who received any CNS prophylactic treatment, such as intrathecal chemotherapy, intraveneous high-dose methotrexate or whole brain irradiation, were excluded to evaluate the natural risk of CNS events in R-CHOP therapy. This study also took particular note of the evaluation of various extranodal involvement sites at presentation.
Materials and Methods
Patients, diagnosis, treatment and inclusion/exclusion criteria. In October 2009, the Bay-area Lymphoma Information Network (Bay-LINK) in Japan, a cooperative study group consisting of the Cancer Institute of the Japanese Foundation for Cancer Research and the Yokohama City University Hematology Group, performed a mail and e-mail survey about CNS involvement in patients with DLBCL. By June 2010, all clinical data had been collected by Bay-LINK.
All patients had been diagnosed with de novo DLBCL and had undergone primary therapy between September 2003 and December 2006. R was approved for the treatment of CD20-positive aggressive B-cell lymphoma in Japan in September 2003 by the Ministry of Health, Labour and Welfare. Patients with distinct forms of DLBCL, such as intravascular lymphoma, primary effusion lymphoma and primary mediastinal large B-cell lymphoma, were excluded from the study. Primary CNS lymphoma and intraocular lymphoma were also excluded in this study. Pathological diagnosis was made by the pathologists in each institution, and no central pathological review was performed. The primary therapy consisted of R-CHOP therapy in all cases. The schedule of the drug dosing was as follows: 50 mg/m2 ADR on day 1, 750 mg/m2 CPA on day 1, 1.4 mg/m2 (maximum 2.0 mg/body) VCR on day 1, 100 mg/body of PSL on days 1–5, and 375 mg/m2 R per cycle. All patients received at least one cycle of R-CHOP therapy with curative intent. The chemoimmunotherapy was performed every 3 weeks. The number of patients who received the therapy at an interval of more than 1 week postponement per cycle or who received ADR or CPA with more than a 20% average dose reduction was investigated. The number of patients who underwent local irradiation as part of their primary therapy was also investigated. Along with factors including age, gender, clinical stage, serum lactate dehydrogenase (LDH) level, performance status, bulky mass, B symptoms, revised International Prognostic Index (R-IPI) and number of extranodal involvement sites, the presence or absence of the following 26 extranodal involvement sites was recorded: orbita, nasal sinus, paranasal sinus, Waldeyer’s ring, salivary gland, thyroid gland, breast, thymus, lung, pleura, stomach, small intestine, colon, peritoneum, liver, pancreas, spleen, kidney, adrenal gland, testis, ovary/uterus, bone, bone marrow, peripheral blood, skin and subcutis.
Patients were included if they were ≥15 years of age and staged using, at minimum, physical examination, computed tomography from neck to pelvis, and bone marrow examination. Each factor of the R-IPI,(18) as well as the presence or absence of bulky mass defined as having a diameter of at least 10 cm, was assessed. The following exclusion criteria were applied: (i) patients who received any CNS prophylaxis during the clinical course; (ii) patients with initial CNS involvement at presentation; (iii) patients with active cancer; and (iv) patients with human immunodeficiency virus infection. This study was approved by the Yokohama City University Hospital Clinical Research Ethics Board. The procedures of the present study were in accordance with the Helsinki Declaration.
Central nervous system disease. Central nervous system disease was diagnosed when malignant cells were detected in cytocentrifuged preparations of cerebrospinal fluid (leptomeningeal type) and/or when an intracranial or spinal mass was detected by radiologic imaging, such as computed tomography or magnetic resonance imaging (parenchymal type). In the present study, epidural spinal cord compression was not considered a CNS disease. Patients with symptoms suggesting CNS disease without cytological or radiological findings were not regarded as having CNS disease. In the present study, CNS disease that occurred during systemic complete remission and during systemic active lymphoma was counted as a “CNS event.”
Statistical analysis. Overall survival (OS) was calculated from the date of initiation of the therapy or, where applicable, from the date of the CNS event to the date of last follow up or death from any cause. Time to CNS event (TTCNS) was calculated from the date of initiation of the therapy to the date of the CNS event. Survival analysis and TTCNS were estimated using the Kaplan–Meier method and compared using the log-rank test. A P-value <0.05 indicated statistical significance. To evaluate the risk factors for CNS events, a univariate analysis was carried out using TTCNS as the end point. A Cox proportional hazards model including all factors with P < 0.1 from the univariate analysis was performed to determine the impact of those factors on the risk of a CNS event. Data were analyzed using the Statistical Package for the Social Sciences (IBM PASW Statistics 18.0, IBM Corporation, Armonk, NY, USA).
Baseline characteristics. Clinical data from 1221 patients with CD20-positive DLBCL were collected from 47 institutions in Japan. The median age was 64 years, ranging from 15 to 91 years. Patient characteristics, including the five risk factors of the International Prognostic Index (IPI), are listed in Table 1. According to the R-IPI, 433 patients (35.2%) were included in the “poor” risk group (≥3 risk factors).
Table 1. Baseline characteristics and therapeutic factors
Characteristic or therapeutic factors
Number of patients (%)
ADR, doxorubicin; CPA, cyclophosphamide; HD, lactate dehydrogenase; IPI, international prognostic index; N, upper limit of normal range; PS, performance status.
Number of R-CHOP courses
Dose reduction (≥20% of ADR)
Dose reduction (≥20% of CPA)
Extended interval per course over 1 week
Therapeutic factors are also summarized in Table 1. R-CHOP therapy was performed in all 1221 patients as the primary treatment. Of the 1221 patients, 910 (74.6%) were treated with 6–8 cycles of therapy. Dose reductions of ADR and CPA >20% were noted in 24.5% and 22.7% of patients, respectively. Two hundred and six patients (17.0%) were treated with an extended interval per course of >1 week. Local irradiation was added in 297 patients (24.4%).
The OS for the entire cohort of 1221 patients is depicted in Figure 1(A). The median observation period in living patients was 47.9 months. The 5-year OS rate was 75.0%. The R-IPI was predictive in identifying the three risk groups (Fig. 1B, P < 0.001). The 5-year OS rate was 95.6% in the “very good” risk group (0 risk factors), 84.0% in the “good” risk group (one or two risk factors) and 53.5% in the “poor” risk group.
Incidence of central nervous system events. In total, 82 CNS events (6.7%) were recorded. More than half of the CNS events were of the parenchymal type (53.7%), followed by the leptomeningeal type (31.7%) and both (14.6%). CNS events occurred during the first complete remission (CR) in 38 patients (46.3%) and in the second or later CR in eight patients (9.8%) as isolated CNS recurrences (Table 2). The remaining 36 CNS events occurred in patients with relapsed or primary refractory status. In the 38 patients with CNS events during the first CR, median TTCNS was 12 months and the types of CNS events were of the parenchymal type in 24 patients (63.1%), the leptomeningeal type in six patients (15.8%) and both in eight patients (21.1%). In the 82 patients with CNS events, death from any cause was recorded in 53 of these patients (64.6%) during the observation period, with most deaths occurring due to lymphoma. Patients with a CNS event showed significantly worse survival compared to patients without a CNS event (Fig. 1C, P < 0.001). The TTCNS curve of the entire cohort is depicted in Figure 2(A). The 5-year probability of a CNS event was 8.4%. The median time interval between the date of initiation of therapy and the CNS event was 9 months (range, 1–55 months). Of the 82 CNS events, 11 (13.4%) occurred after 36 months of observation and were considered late CNS events, of which five were of the parenchymal type, three of the leptomeningeal type and three were both. The late CNS events occurred during the first CR in six cases, the second later CR in four cases, and non-CR in one patient, who had already relapsed systematically. Comparing the TTCNS curves in patients with or without a dose reduction of more than 20% of ADR, the TTCNS observation was significantly inferior in patients with a dose reduction (Fig. 2B, P = 0.012). The difference was marginally significant in patients with or without a dose reduction of more than 20% of CPA (Fig. 2C, P = 0.057). Furthermore, the extended interval per course over 1 week had no influence on the TTCNS (data not shown).
Table 2. Type of CNS event, systemic status, and outcome
CNS, central nervous system; CR, complete response; N, number.
Type of CNS event
Second or more CR
Outcome at the latest contact
Death from lymphoma
Death from other causes
Risk factors for central nervous system event. In the univariate analysis, 36 risk factors were evaluated for increased risk of a CNS event. Twenty-two risk factors with P-value <0.1 are shown in Table 3. Among the general risk parameters, occurrence of a CNS event was associated with age over 60 years, advanced stage, elevated LDH, more than twofold elevated LDH, poor Eastern Cooperative Oncology Group performance status, presence of B symptoms, poor risk group according to R-IPI, and the presence of two or more extranodal involvement sites with P-value <0.05. Among the local risk parameters, involvement of the paranasal sinus, Waldeyer’s ring, salivary gland, breast, pleura, peritoneum, spleen, kidney, adrenal gland, bone, bone marrow and peripheral blood were also associated with increased risk of a CNS event with P-value <0.05.
Table 3. Factors associated with increase probability of CNS event
CNS/all (n = 82/1221)
CI, confidence interval; CNS, central nervous system; EN, extranodal involvement sites; H; high risk; HI, high–intermediate risk; IPI, international prognosis index; L, low risk; LDH, lactate dehydrogenase; LI, low–intermediate risk; N, upper limit of normal range; NS, not significant; PS, performance status; RR, relative risk.
Multivariate Cox regression analysis including the 22 risk factors with P-value <0.1 by univariate analysis identified involvement of breast (relative risk [RR] 10.5), adrenal gland (RR 4.6) and bone (RR 2.0) as the risk factors for CNS events (Table 3). Age over 60 years was also identified as a risk factor for CNS events (RR 2.1). TTCNS curves significantly differed between patients with and without any (one or more) of the three CNS risk factors (Fig. 2D, P < 0.001). The 5-year probability of a CNS event in patients with and without CNS risk factors was 22.2% and 6.9%, respectively.
Survival after central nervous system event. Figure 3(A) shows the OS after a CNS event in 82 patients. The 2-year survival rate was 27.1%, and the 50% survival duration was 6.1 months. Among the three types of CNS events, no significant differences were observed between any two types (Fig. 3B). According to the systemic lymphoma status at the time of the CNS event, patients who experienced CNS events in the first CR did not show superior survival as compared with the others (Fig. 3C, P = 0.16). Moreover, patients with CNS events in any CR also did not show superior survival as compared with the others (Fig. 3D, P = 0.12).
In the entire cohort of 1221 adult patients with DLBCL in the R era, CNS events occurred in 6.7% of patients, and the 5-year cumulative incidence of CNS events was 8.4%. Several reports have focused on CNS events in patient cohorts of more than 1000;(19–21) however, these studies have included patients with heterogeneous histological diagnoses. The present study is the largest to investigate one uniform histological diagnosis and uniform treatment. The OS of the patients according to the R-IPI was similar to that of the original report,(18) confirming that DLBCL patients in this study had a regular disease risk. Previous reports have demonstrated a CNS event incidence of approximately 5% in the pre-R era.(13) Although our results from the R era cannot be directly compared with those findings, we presume that the incidence of a CNS event has not decreased in the R era. Intravenous R enters the CNS only at very low levels and contributes only to the improved prognosis of systemic lymphoma and not to the prevention of CNS events.(22) Therefore, CNS events remain an important research issue in the R era. In this study, we retrospectively collected data from multiple centers in Japan for patients without CNS prophylaxis. The role of CNS prophylaxis in patients with DLBCL is not clear, except in patients with testicular involvement in whom the CNS prophylactic strategy is already justified.(23,24) It was anticipated that the subject and the method of CNS prophylaxis might vary among institutions. Although the number of patients who received CNS prophylaxis during the study period in the 47 institutions was unclear, we believe this is a reasonable method for evaluating the risk of CNS events in R-CHOP therapy, even if the possibility of excluding high-risk patients from the analysis exists.
The 2-year OS after a CNS event in 82 patients was 27.1%. Of the 54 deaths recorded during the observation period, most (52/54 patients) occurred as a result of the progression of lymphoma. Although most reported cases of CNS events are of the leptomeningeal type, the parenchymal type has been reported to be predominant in patients with DLBCL.(25) Similarly, more than half of the cases of CNS events in the present study (44/82 patients) were of the parenchymal type. Lymphoma cells might penetrate the blood brain barrier into the CNS through the blood stream. Among patients who experienced CNS events in the first CR, approximately two-thirds of CNS events were of the parenchymal type (24/38). CNS prophylaxis, such as with high-dose methotrexate, might be justified for preventing parenchymal type CNS events in the first CR. However, no significant differences were observed between any pair of the three types of CNS events in our study.
There are several reports concerning CNS prophylaxis in the pre-R era.(5,7–12) The efficacy of CNS prophylaxis is controversial, mainly because the incidence rate of CNS events is rather low, at approximately 5%. Most reports have included varied lymphoma histology, and their inclusion or exclusion of patients who received CNS prophylaxis has varied. Five recent reports(21,25–28) evaluate the effect of R on CNS events, including two publications by our group.(26,28) Interestingly, R is found to have no influence on the prevention of CNS events in two studies involving patients without CNS prophylaxis,(25,28) but a positive effect is noted in three studies involving patients who received CNS prophylaxis.(21,26,27)
To resolve these discrepancies, we retrospectively analyzed a large cohort of patients with DLBCL without CNS prophylaxis in the R era. Each extranodal involvement site was also evaluated. In the Cox proportional hazards model, three extranodal involvement sites and age over 60 years were identified as risk factors. It is unrealistic to propose that all patients aged over 60 years receive CNS prophylactic treatment; therefore, the realistic risk factors extracted from the current study are breast involvement, adrenal gland involvement and bone involvement. The testis is well known as a high-risk involvement site,23,24 and prophylactic irradiation to the contralateral testis after surgery to remove the affected testis as well as CNS prophylactic intrathecal chemotherapy are recommended in patients with testicular lymphoma, according to National Comprehensive Cancer Network guidelines.(29) Patients with testicular involvement were likely to have undergone CNS prophylaxis, as described above, and, therefore, were not included as subjects in the present study. In contrast, the breast has only recently been reported as a high-risk involvement site.(30) Therefore, most patients with breast involvement from 2003–2006 were likely to have been treated without CNS prophylaxis and were included as subjects of the present study, resulting in a 35% incidence (6/17 patients) of CNS events (RR 10.5).
In the present study, the frequency of CNS high-risk patients with at least one of the three extranodal involvement sites was 10.9% (133/1221 patients). According to previous reports, patients with testicular involvement account for 1–2% of all non-Hodgkin lymphoma cases.(24) Consequently, we defined CNS high-risk patients as those with the following four involvement sites, who might account for approximately 12% of all patients with DLBCL: testis, breast, adrenal gland and bone. Notably, CNS events occurred in only approximately 22% of CNS high-risk patients. In the four extranodal sites mentioned above, the efficacy of CNS prophylaxis has been confirmed only in patients with testicular involvement and has not yet been established in the others. The best method of prophylaxis for each involvement site remains to be elucidated. In addition, the administration of an adequate dose of key drugs, such as ADR and CPA, might be another important factor for preventing CNS events, despite the use of R.
We identified CNS high-risk patients with DLBCL in the R era. In the future, a randomized controlled study of CNS high-risk patients to evaluate the role of prophylaxis is highly anticipated. A new method for the early diagnosis of CNS involvement at the time of presentation should also be established.
We are grateful to Yoshio Saburi (Oita Prefectural Hospital), Haruhisa Nagoshi (St. Marianna University School of Medicine Yokohama City Seibu Hospital), Jun Ishikawa (Osaka Medical Center for Cancer and Cardiovascular Diseases), Takaaki Miyake (Shimane University Hospital), Masaaki Noda (Hiroshima City Hospital), Takashi Okamura (Kurume University Hospital), Jun Takizawa (Niigata University Medical and Dental Hospital), Shingo Yano (Jikei University Hospital), Morio Matsumoto (Nishigunma National Hospital), Masanobu Nakata (Sapporo Hokuyu Hospital), Norio Yokose (Nippon Medical School Chiba Hokusoh Hospital), Masayuki Hino (Osaka City University Hospital), Takahiko Utsumi (Shiga Medical Center for Adults), Tomufumi Yano (Okayama Rosai Hospital), Nobuhiko Uoshima (Matsushita Memorial Hospital), Haruko Tashiro (Teikyo University Hospital), Yuji Kanisawa (Oji General Hospital), Yoshinori Tanaka (Yamaguchi University Hospital), Yoshimasa Kura (Kasukabe Municipal Hospital), Michiaki Koike (Juntendo University Shizuoka Hospita), Shiro Matsuura (Shizuoka Red Cross Hospital), Gou Aoki (Keiju Medical Center), Juichi Tanabe (Fujieda Municipal General Hospital), Sadaya Matano (Tonami General Hospital), Masanori Kume (Hiraka General Hospital), Tatsuyuki Hayashi (Tokyo Metropolitan Police Hospital), Kunio Hayashi (Hirakata Kohsai Hospital) and Takamasa Hayashi (Tenri Hospital). This work was supported in part by the Foundation for Promotion of Cancer Research in Japan.
The authors declare no competing financial interests.