Int J Hematol (2009) 89:577–583 DOI 10.1007/s12185-009-0289-2
ORIGINAL ARTICLE
Diffuse large B-cell lymphoma with central nervous system relapse: prognosis and risk factors according to retrospective analysis from a single-center experience Yutaka Shimazu Æ Kenji Notohara Æ Yasunori Ueda
Received: 7 November 2008 / Revised: 18 February 2009 / Accepted: 24 February 2009 / Published online: 9 April 2009 Ó The Japanese Society of Hematology 2009
Abstract The introduction of rituximab for diffuse large B-cell lymphoma (DLBCL) has improved the disease’s overall prognosis. However, relapse in the central nervous system (CNS) is still an issue. We investigated the prognosis and risk factors of CNS recurrence in DLBCL. A total of 403 patients who were diagnosed with DLBCL without CNS involvement between January 1996 and April 2007 at our institution were included in the study. Subsequently, 42 experienced CNS relapse. Clinical information was gathered by chart review. The median disease-free interval to CNS relapse was 625 days. The mean survival periods after relapse in the cases with CNS and extra-CNS involvement were 513 and 1,615 days, respectively (P = 0.0004). Multivariate analysis identified age[60 years (P = 0.031), involvement in two or more extranodal sites (P = 0.040), bone marrow involvement (P = 0.036), an elevated serum lactate dehydrogenase (LDH) level (P = 0.016), and treatment without rituximab before CNS relapse (P = 0.027) as independent predictors of CNS relapse. We have This work was done in the Department of Hematology/Oncology, Kurashiki Central Hospital, Okayama. Part of this study was presented at the 49th Annual Conference of the American Society of Hematology, 2008 (Abstract No. 955341). Y. Shimazu Y. Ueda Department of Hematology/Oncology, Kurashiki Central Hospital, Okayama, Japan Y. Shimazu (&) Department of Hematology/Oncology, Graduate School of Medicine, Kyoto University, 54 Shogoin Kawahara-cho, Sakyo-ku, Kyoto 606-8507, Japan e-mail:
[email protected] K. Notohara Department of Pathology, Kurashiki Central Hospital, Okayama, Japan
shown that cases of DLBCL occurring in advanced age, involving two or more extranodal sites or the bone marrow, or showing an elevation of LDH have a higher risk of CNS relapse. Rituximab may prevent CNS relapse by reducing the recurrence of DLBCL at all sites. An effective CNS prophylaxis strategy should be determined according to the risk assessment of CNS relapse. Keywords CNS relapse Diffuse large B-cell lymphoma Rituximab Risk factors Retrospective study
1 Introduction The introduction of rituximab therapy for diffuse large Bcell lymphoma (DLBCL) has dramatically improved the overall prognosis of this disease. The addition of rituximab to the CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) regimen has been shown to increase the complete remission (CR) rate and overall survival in elderly patients with DLBCL, compared to the standard CHOP regimen [1–3]. However, the prognosis of cases with relapse in the CNS is still very poor, and thus it is crucial that a protocol be established for preventing and managing CNS involvement [4–7]. The effectiveness of rituximab against CNS lymphoma is a matter of controversy [8]. There are, for example, no studies presenting clear evidence that rituximab prevents CNS relapse [9, 10]. However, several studies have shown that intrathecal chemotherapy prophylaxis may reduce the incidence of CNS relapse [11–13]. Combination treatment that includes high-dose methotrexate chemotherapy might be effective, but the incidence of relapse in the CNS is not high enough to warrant the use of a CNS prophylaxis strategy in all DLBCL patients [14]. Previous studies have focused on
123
578
Y. Shimazu et al.
identifying risk factors for CNS relapse in DLBCL in order to determine the higher-risk groups for whom CNS prophylaxis might be beneficial [4, 5, 14–17]. In the present study, we investigated the risk factors and prognosis of CNS relapse in DLBCL by retrospective analysis of the cases at our institute. 2 Subjects and methods From January 1996 to April 2007, 437 patients were diagnosed with DLBCL at Kurashiki Central Hospital. Among these 437 cases, 2 cases were excluded because they had already been treated at other institutes, and 32 cases were excluded due to CNS involvement at the time of diagnosis. Therefore, 403 cases were analyzed in the following study (Table 1). All the pathological diagnoses were confirmed by the pathologist at our institution. A mixed component of low-grade lymphoma was observed in 37 cases, 3 of which showed relapse in the CNS. We included these 37 cases in our analysis only after the diagnosis of transformation to DLBCL had been determined. More than half of the total patient group were men and more than 75% of patients were older than 60 years (the median age was 71 years). To estimate the clinical stages, computed tomography (CT) and/or gallium scintigraphy and/or positron emission tomography (PET)-CT scan, endoscopy and bone marrow examination were performed in all cases. More than half of the cases had stage III or IV disease, and more than 30% of cases had involvement in two or more extranodal sites. Bone marrow involvement was present in about 20% of the cases. Using
the international prognostic index (IPI), 36, 14, 23 and 27% of the cases were classified as cases of low, low-intermediate, high-intermediate, and high risk, respectively. Thus, about half of the cases were classified as high-intermediate or high-risk cases. Bone marrow involvement was identified in 76 cases, and the specimen was insufficient to assay for the presence of bone marrow involvement in 12 cases. The evaluation of complete remission after the treatment was made using CT or PET-CT scan. The rate of complete remission was about 90% in patients under 70 years old and about 81% in patients over 70 years old. 2.1 Treatment regimens In our institute, before September 2003, 161 patients were treated with a regimen based on CHOP, and after September 2003, 242 patients were treated with a regimen based on CHOP plus rituximab. Among the 200 patients who were less than 70 years old, 127 patients received rituximab before CNS relapse. Among the 203 patients who were more than 70 years old, 111 patients received rituximab. We also used rituximab for the treatment of the cases that relapsed after September 2003. For patients with stage I disease, we used three or four courses of a CHOPbased regimen plus eight cycles of rituximab. Patients with stage II disease were treated with six courses of a CHOPbased regimen plus eight cycles of rituximab. Depending on the involved sites of stage I or II disease, radiation therapy was also added. For stage III or IV disease, patients were treated with six to eight courses of a CHOP regimen plus eight cycles of rituximab. According to the patient
Table 1 Univariate analysis for CNS relapse Total
No CNS relapse
CNS relapse
Cases
403
361
42
Sex (male/female)
215/188
192/169
23/19
0.7380
Age (median ± 2SD) (range)
71 ± 13 (17–92)
70 ± 13 (17–92)
72 ± 9 (47–85)
0.0344
Age less than 60/more than 60a
97/306
90/271
7/35
Stage (I–II/III–IV)
a
P value
65 ? 111/67 ? 160
61 ? 103/62 ? 135
4 ? 8/5 ? 25
0.0005
LDH (not elevated/elevated)a
188/215
176/185
12/30
0.0001
Performance status (0–1/2–4)a
304/99
273/88
31/11
0.0729
Extra-nodal sites (0–1/C2)a
282/121
258/103
24/18
0.0011
IPI (0/1/2/3/4/5)a
36/109/57/93/72/36
35/103/48/86/57/32
1/6/9/7/15/4
0.0000
Bone marrow involvement (?/-/NE)a
76/315/12
62/288/11
14/27/1
0.0003
Prophylactic intrathecal MTX (?/-)b Rituximab (before CNS relapse) (?/-)b
18/385 238/165
17/344 218/143
1/41 20/22
0.4780 0.0764
Follow-up (days, median) (range)
632 (0–4217)
974 (0–4217)
552 (75–4194)
0.6291
No relapse/relapse/therapy resistance
185/159/59
185/117/59
0/42/0
LDH lactate dehydrogenase, IPI the international prognostic index, MTX methotrexate, CNS central nervous system, NE not evaluated a
At diagnosis
b
Before CNS relapse
123
Risk factors for CNS relapse in DLBCL Table 2 Chemotherapy regimen for DLBCL
579
Patients under 69 years old:165 cases Rituximab ? CHOP regimen (doxorubicin 50 mg/m2 on day 1, cyclophosphamide 750 mg/m2 on day 1, vincristine 1.4 mg/m2 on day 1 and prednisone 100 mg on days 1–5) Patients from 70 to 79 years old:142 cases Rituximab ? THP-COP regimen (pirarubicin 30 mg/m2 instead of doxorubicin on day 1, cyclophosphamide 500 mg/m2 on day 1, vincristine 1 mg/m2 on day 1 and prednisone 30 mg/m2 on days 1– 5) Patients over 80 years old: 46 cases Rituximab ? age-adjusted THP-COP regimen (pirarubicin 21 mg/m2 instead of doxorubicin on day 1, cyclophosphamide 350 mg/m2 on day 1, vincristine 1 mg/m2 on day 1 and prednisone 30 mg/m2 on days 1– 5)
age, chemotherapy regimen was modified [18] as shown in Table 2. The chemotherapy dose was reduced or changed in 50 cases due to poor performance: 19 of these 50 patients were less than 69 years old, 19 were between 70 and 79, and 12 were more than 80 years. In six cases, the cycles of chemotherapy were increased to maintain the total dose of chemotherapy; three of these patients were more than 70 years old and three were less than 70 years. Indicators for CNS prophylaxis included the involvement of nasal sinuses, testis or vertebra, and four courses of intrathecal methotrexate were administered in 18 cases [8, 13, 19]. No additional therapy was added to prevent CNS relapse. Once relapse in the CNS was detected, the patients were treated with combination chemotherapy, i.e., systemic salvage chemotherapy such as a DeVIC (carboplatin, etoposide, ifosfamide and dexamethasone) regimen in combination with rituximab and intrathecal methotrexate and/or high-dose methotrexate and/or radiation [4–6, 20, 21]. When the patients relapsed in sites other than the CNS, they were treated with systemic salvage chemotherapy such as a DeVIC regimen or EPOCH regimen (doxorubicin, etoposide, vincristine, cyclophosphamide, and prednisone) in combination with rituximab, followed by autologous transplantation when indicated [22–24]. As much as 21 patients received autologous transplantation. 2.2 Diagnosis of CNS relapse Patients with a CNS lesion at the initial time of diagnosis were not included in cases with CNS relapse [19]. CNS relapse was observed both during and after treatment. Lumbar puncture and brain MRI were indicated for the cases in which intrathecal methotrexate had been administered or for the cases with suspected neurological abnormalities. The diagnosis of CNS relapse was based on radiological findings, brain biopsy histology and spinal tap cytology. Pathological procedures were carried out when possible, but, in a few cases, the diagnosis of CNS relapse
was based only on radiological findings and clinical symptoms. 2.3 Statistical analysis In this retrospective analysis, the primary end point was the time to CNS relapse, and this period was defined as the time from the diagnosis to disease progression with CNS involvement. The secondary end point was the survival time after CNS relapse, and this period was defined as the time from CNS relapse until death from any cause. The cumulative incidence of CNS relapse was calculated by treating deaths from any cause as a competing risk [25]. To identify prognostic factors for CNS relapse, a logistic regression analysis was carried out to assess the effect of pretreatment prognostic factors (age [ 60 years, sex, involvement in two or more extranodal sites, stage, Eastern Cooperative Oncology Group (ECOG) performance status (PS) [ 1, increased serum level of lactate dehydrogenase (LDH), bone marrow involvement, rituximab use and prophylactic intrathecal methotrexate use) [26, 27]. The strength of prognostic factors was estimated by calculating relative risks (RRs) and the corresponding 95% confidence intervals (CIs). All P values of\0.05 were considered to be significant. The patient characteristics and the distribution of time intervals to relapse were compared by v2 tests. Survival curves were calculated by the Kaplan–Meier method. All calculations were made using the program SPSS/PC?, version 11.0. This study was approved by our institutional review board and conducted by following the Declaration of Helsinki.
3 Results The clinical characteristics of the 403 patients are shown in Table 1. During the median follow-up period of 632 days,
123
580
Y. Shimazu et al.
59 patients were chemotherapy-resistant (all chemotherapy failed). Among the 344 patients in whom chemotherapy was initially successful, 159 patients experienced relapse of DLBCL, and 42 of these 159 patients experienced CNS relapse. No relapse was observed in 185 cases. Rituximab was administered in about 60% of cases. Among the 42 cases with CNS relapse, brain parenchymal involvement and meningeal relapse was diagnosed in 32 and 10 cases, respectively. CNS was the only site of relapse in 28 cases, whereas both CNS and extra-CNS sites were involved in 14 cases. Among the 28 cases in which CNS was the only site of relapse, rituximab was used in 14 cases and not used in 14 cases. On the other hand, among the 14 cases in which CNS and extra-CNS sites were involved, rituximab was used in 8 cases and not used in 6 cases. The median interval between the diagnosis and CNS recurrence was 625 days. As much as 22 cases had CNS relapse within 1 year. The occurrence of CNS relapse at one year was 6.47% (the 95% confidence interval was 6– 7.14%; Fig. 1). The median interval of CNS relapse was 655 days in patients who received rituximab and 598 days for those who did not. No difference in relapse intervals was seen between CNS and extra-CNS sites (Table 4). Among the 18 patients who received prophylactic intrathecal administration of methotrexate, only one case developed CNS relapse afterwards. The chemotherapy dose was reduced or changed in 50 cases due to poor performance: among those patients who received dose reduction chemotherapy, and only 4 cases relapsed in the CNS. As shown in Table 1, univariate analysis identified six factors as predictors of relapse in the CNS: age [ 60 (P = 0.028), advanced disease stage (stage III–IV; P = 0.0005), an elevated serum level of LDH (P = 0.0001), involvement in two or more extranodal sites (P = 0.0011),
bone marrow involvement (P = 0.0003), and an increased IPI score (P \ 0.0001). The other four factors, use of rituximab before CNS relapse, sex, ECOG performance status [ 1, and prophylactic intrathecal methotrexate use, were not predictive of CNS relapse by univariate analysis. Multivariate analysis identified age [ 60 (P = 0.031: odds ratio = 2.501), involvement in two or more extranodal sites (P = 0.040: odds ratio = 2.018), bone marrow involvement (P = 0.036: odds ratio = 2.099), an elevated serum LDH level (P = 0.016: odds ratio = 2.406) and treatment without rituximab before CNS relapse (P = 0.027: odds ratio = 0.485) as independent predictors of CNS relapse (Table 3). The overall survival rate of cases with CNS relapse was significantly lower than that of non-relapsed cases, but no difference was observed between the cases with CNS and extra-CNS relapses (Fig. 2). The overall survival of chemotherapy-resistant patients was significantly worse
Table 3 Multivariate analysis for CNS relapse P value Hazard ratio (95%CI) Sex (male/female)
0.227
Age [ 60
0.031
Stage (I–II/III–IV)a
0.230
Performance status (0–1/2–4)a a
Extra-nodal sites (0–1/C2)
2.501 (1.087–5.754)
0.735 0.040
2.018 (1.033–3.942)
Bone marrow involvement (?/-)a 0.036 LDH (not elevated/elevated)a 0.016
2.099 (1.049–4.200) 2.406 (1.181–4.903)
Prophylactic IT (?/-)b
0.571
Rituximab (?/-)b
0.027
0.485 (0.255–0.922)
IT intrathecal methotrexate a
At diagnosis
b
Before CNS relapse
1.0
1.0
.9
(n=403)
No relapse
.8
.8
overall survival
hazard
.7 .6 .5 .4
.6
Relapse other than CNS
.4
.3 .2
CNS relapse
.2
.1
Chemotherapy resistance
0.0 0
500
1000
1500
2000
days
Fig. 1 Cumulative incidence of CNS relapse in DLBCL
123
2500
0.0 0
500
1500
1000
2000
days
Fig. 2 Overall survival of patients with DLBCL (n = 403)
2500
Risk factors for CNS relapse in DLBCL
581
extra-CNS relapse (Fig. 4b). However, rituximab did not improve the overall survival of patients with CNS relapse (Fig. 4c). There was no significant difference in salvage chemotherapy between the group treated with and that treated without rituximab, but more cases which relapsed within 1 year were included in the patients treated without rituximab. The patients’ backgrounds were compared between the group receiving and that not receiving rituximab. No differences in age, disease stage or extranodal sites were observed, but the patients not receiving rituximab had a higher serum lactate dehydrogenase level and worse performance status. Among the cases that had relapsed at sites other than the CNS at more than 3 years after the initial diagnosis, ten cases were treated with rituximab. Three of these ten cases relapsed in the CNS after the salvage therapy.
compared to the other three groups. After the relapse, patients with CNS relapse had a worse survival rate compared to those with only extra-CNS involvement (Fig. 3), and only 8 out of the 42 patients with CNS relapse were alive at the time of analysis. The introduction of rituximab therapy for DLBCL significantly improved its overall prognosis (Fig. 4a). The benefit of using rituximab could also be seen in cases with 1.0
p=0.0004
.6
Relapse other than CNS
(n=117)
.4
4 Discussion
(n=42)
Relapse in CNS .2
0.0
0
500
1000
1500
2000
In the present analysis, 42 cases relapsed in the CNS among 403 cases of DLBCL. The incidence of CNS relapse reached about 10% and was much higher than those reported in previous studies [4–6, 14–16]. This discrepancy may be explained by the higher proportion of patients in the high risk group and longer follow-up in our study. In fact, more than half of our cases were in the
2500
days
Fig. 3 Overall survival after relapse of patients with DLBCL (n = 159)
(a) 1.0
p
