ABSTRACT
162 patients with relapsed secondary central nervous system lymphoma (R‐SCNSL) (median age, 65 years; male, 59.9%) including central nervous system (CNS)‐only (n = 120) and concomitant CNS/systemic relapse (n = 42) were retrospectively analyzed. Overall, 21.9% of patients were classified as high risk according to the CNS International Prognostic Index (CNS‐IPI). Several biological and clinical features were significantly associated with leptomeningeal involvement, including double‐ or triple‐hit (DHL/THL) status, MYC rearrangement, negative BCL6 expression by IHC, bone marrow involvement, and concomitant R‐SCNSL. Multivariable analysis showed that leptomeningeal involvement independently predicted inferior OS and was associated with a 98% increase in the hazard of death compared with parenchymal relapse (HR = 1.98, 95% CI: 1.20–3.26, p = 0.008). Based on anatomical localization, R‐SCNSL was classified into four subtypes: parenchymal‐only involvement (parenchymal‐CNS [P‐CNS], 68/42%) and parenchymal involvement plus systemic relapse (parenchymal‐concomitant [P‐concomitant], 17/10.5%); and leptomeningeal with or without parenchymal involvement (leptomeningeal‐CNS [LM‐CNS], 52/32.1%) and leptomeningeal with systemic relapse (leptomeningeal‐concomitant [LM‐concomitant], 25/15.4%). This anatomical classification significantly impacted OS and PFS (p < 0.001). Two‐year OS and PFS were 58.2% and 29.1% for P‐CNS, 32.4% and 17.7% for P‐concomitant, 22% and 13.9% for LM‐CNS, and 7.1% and 0% for LM‐concomitant, respectively. ASCT showed a trend toward improved survival among patients with a response (CR/PR) in a 4‐month landmark analysis. These findings support the clinical application of the anatomical classification in the management of R‐SCNSL.
Keywords: CNS relapse, diffuse large B‐cell lymphoma, high grade B‐cell lymphoma, leptomeningeal involvement, parenchymal involvement
1. Introduction
Secondary central nervous system lymphoma (SCNSL) refers to central nervous system (CNS) dissemination and involvement at the time of initial diagnosis (de novo SCNSL, DN‐SCNSL) or at relapse (relapsed SCNSL, R‐SCNSL). R‐SCNSL can present as CNS‐only or with concomitant CNS and systemic disease. [1, 2, 3, 4, 5, 6, 7, 8].
In the largest clinical trial of diffuse large B‐cell lymphoma (DLBCL) with secondary central nervous system involvement (MARIETTA trial), DN‐SCNSL, CNS‐only R‐SCNSL, and concomitant R‐SCNSL represented 43%, 20%, and 37% of the 75 patients enrolled, respectively. [5] The neuroanatomical localization of SCNSL included intracerebral (45%), leptomeningeal (11%), intracerebral/leptomeningeal (17%), intracerebral/ocular (13%), intracerebral/leptomeningeal/ocular (8%), ocular (3%), and spinal cord (3%) involvement. Non‐germinal center B‐cell‐like (non‐GCB) DLBCL was observed in 59% of cases, and neurological manifestations were associated with the specific sites of CNS involvement. [5] In the largest international study on SCNSL, 522 patients with DN‐SCNSL were reported, of whom 44.7% had parenchymal (P); and 40.4% leptomeningeal (LM) involvement. Among 588 patients with R‐SCNSL, 66.2% had CNS‐only relapse and 33.8% had concomitant CNS/systemic relapse, with 56.8% P involvement and 22.6% LM involvement. [7].
SCNSL is relatively uncommon, identified in 1.05% of aggressive lymphomas at initial diagnosis [9] and in 2.2%–2.8% of relapsed cases. [1, 9] Prior studies have reported a CNS relapse rate of 3%–5% among patients with systemic DLBCL. [6, 10, 11, 12] DN‐SCNSL and R‐SCNSL appear to differ in terms of clinical and biological features, as well as in the selection of therapeutic agents for initial treatment. Notably, anatomical involvement is more variable in R‐SCNSL than in DN‐SCNSL. Here, we report the findings from our comprehensive anatomical localization–based analysis of R‐SCNSL and propose an anatomical classification system for clinical application.
2. Methods
2.1. Study Design
Patients with R‐SCNSL aggressive B‐cell lymphomas treated at the Mayo Clinic between 2002 and 2025 were identified via medical records and the Mayo Clinic Lymphoma Database and analyzed retrospectively. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Mayo Clinic Institutional Review Board (IRB). Extracted data included clinicopathologic features, radiologic findings, treatment details, and survival outcomes.
R‐SCNSL was defined as a CNS relapse, with or without systemic involvement, following frontline therapy for a systemic lymphoma that had no CNS involvement at initial diagnosis. The aggressive B‐cell lymphoma pathologic subtypes were determined based on histopathologic and immunophenotypic evaluation at the initial diagnosis of systemic lymphoma, in accordance with WHO criteria. Four subtypes were included in this analysis: germinal center B‐cell‐like (GCB) and non‐GCB DLBCL (defined by the Hans algorithm); high‐grade B‐cell lymphoma, not otherwise specified (HGBCL‐NOS); and double‐hit or triple‐hit lymphomas (DHL/THL, characterized by MYC and BCL2 and/or BCL6 rearrangements). Secondary CNS involvement was classified as parenchymal or leptomeningeal based on magnetic resonance imaging (MRI) findings, brain biopsy and cerebrospinal fluid (CSF) analysis. Leptomeningeal involvement was defined as either positive CSF cytology or radiographic evidence of leptomeningeal disease on MRI. In the parenchymal‐only group (n = 85), 70.6% (60/85) had CSF analysis, and 57.6% (49/85) had brain biopsy. In the leptomeningeal‐only group (n = 61), 96.7% (59/61) had a diagnostic CSF analysis; for the two patients without CSF analysis, lumbar puncture was clinically contraindicated, and diagnosis was based on MRI and clinical findings. In the combined‐compartment group (n = 16), all patients (100%, 16/16) had CSF analysis, and 31.3% (5/16) had brain biopsy. Systemic relapse was determined by positron emission tomography/computed tomography (PET‐CT) or computed tomography (CT) imaging.
Cell of origin status of DLBCL was determined at initial diagnosis using the Hans algorithm. DHL/THL status was assessed by fluorescence in situ hybridization (FISH) analysis for MYC, BCL2, and BCL6 rearrangements at initial diagnosis. MYD88 L265P mutation status was determined from data obtained during routine clinical testing. In our cohort, this analysis was performed for patients with non‐GCB DLBCL using PCR‐based methods on lymphoma tissue obtained at diagnosis or relapse, depending on sample availability. The predominant technique used was allele‐specific quantitative PCR. This assay has a sensitivity with a lower limit of detection for the mutant allele fraction of 0.5% within a background of wild‐type DNA. BCL6 expression was considered positive if immunohistochemistry (IHC) demonstrated staining in ≥ 30% of lymphoma cells.
Treatment regimens for 162 patients with CNS relapse were categorized by intensity. The treatment intensity groups were defined as follows: Low‐intensity regimens consisted of high‐dose methotrexate (HD‐MTX) with or without rituximab (R). Moderate‐intensity therapy was defined as the MRT regimen (methotrexate, rituximab, and temozolomide). High‐intensity therapy included clinically distinct regimens such as MATRix (methotrexate, cytarabine, thiotepa, and rituximab) and R‐ICE (rituximab, ifosfamide, carboplatin, and etoposide). Separate high‐intensity regimens were evaluated in univariable analyses; however, due to limited sample size, these regimens were grouped for multivariable modeling. The “Other” category included heterogeneous regimens not classifiable within predefined treatment‐intensity categories, including intrathecal chemotherapy (intrathecal methotrexate alone or with simple combinations, or triple intrathecal therapy), non‐standard or intermediate‐dose methotrexate‐based regimens, heterogeneous multi‐agent systemic chemotherapy, radiation‐based therapies (whole‐brain radiotherapy or focal stereotactic radiosurgery), targeted therapy (e.g., ibrutinib), site‐specific or institutional protocols (e.g., Ferreri or Vanderbilt protocols), and observation or no treatment.
2.2. Statistical Analysis
Continuous variables were summarized as mean (standard deviation [SD]) and median (range) while categorical variables were reported as frequency (percentage). Continuous variables were compared between groups using Wilcoxon rank‐sum test, and categorical variables were compared using Chi‐squared test or Fisher's exact test, as appropriate. The Kaplan–Meier method was used to estimate overall survival (OS) and progression‐free survival (PFS) after the CNS relapse. Cox proportional hazards regression models were used to identify potential factors associated with OS and PFS. All tests were two‐sided, with p‐values < 0.05 considered statistically significant. Statistical analyses were performed using R version 4.4.1 (R Core Team, 2024)’. Multivariable Cox regression models considered variables that were statistically significant in univariable analyses. Pathologic subtype and MYC rearrangement were highly collinear; therefore, pathologic subtype was retained. BCL6 expression and bone marrow involvement were associated with CNS compartment and, due to collinearity and limited events, were not included. The final model included pathologic subtype, SCNSL classification, CNS compartment, treatment intensity, and ASCT after relapse.
3. Results
3.1. Patient Demographics, Clinical and Pathologic Features, and Frontline Therapeutic Interventions
The cohort included 162 patients with R‐SCNSL, comprising 120 patients with CNS‐only relapse and 42 patients with concomitant CNS/systemic relapse (Supporting Information S1: Table 1). The median age at CNS relapse was 65 years, with 67% (108/162) of patients older than 60. The cohort was 60% male and 87.7% (142/162) Caucasian.
At initial diagnosis, 81.8% (130/159) of patients had advanced‐stage disease (stage III/IV). Extranodal involvement was noted in 69.2% (110/159), with 26.5% (43/162) involving more than one site; specifically, 8.0% (13/162) had kidney/adrenal and 23.8% (38/160) had bone marrow involvement. Regarding risk stratification, 55.2% (58/105) were in high‐intermediate or high‐risk International Prognostic Index (IPI) categories, while 21.9% (23/105) fell into the high‐risk CNS International Prognostic Index (CNS‐IPI) category.
Pathologic subtypes included non‐GCB DLBCL (40.7%; 66/162), GCB DLBCL (27.2%; 44/162), DHL/THL (26.5%; 43/162), and HGBCL‐NOS (5.6%; 9/162). Additionally, 4.9% (8/162) represented transformation from prior indolent lymphoma, and BCL6 expression was identified in 75.2% (109/145).
Variables between the CNS‐only and concomitant CNS/systemic groups were not significantly different, except for age at relapse (> 65 years), elevated LDH at initial diagnosis, MYD88 L265P mutation status, and the distribution of treatment intensity (Supporting Information S1: Table 1). Elevated LDH was more frequent in the concomitant CNS/systemic group compared to the CNS‐only group (96.2% vs. 66.7%, p = 0.002). The MYD88 L265P mutation was assessed in lymphoma tissue from 27 patients with non‐GCB DLBCL. Of these 13 mutation‐positive cases, 9 samples were from CNS tissue and 4 were from non‐CNS sites. The prevalence was significantly higher in the CNS‐only group than in the concomitant group (61.9% vs. 0%; p = 0.016). All 13 patients who tested positive had relapse in the CNS parenchyma, whereas eight of 14 patients who tested negative also had CNS parenchymal relapse.
For first‐line systemic treatment, 72.2% (117/162) received rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R‐CHOP), 10.5% (17/162) received rituximab plus etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin (R‐EPOCH), one patient received rituximab plus cyclophosphamide, vincristine, doxorubicin, methotrexate/ifosfamide, etoposide, and cytarabine (CODOX‐M/IVAC), and 16.0% (26/162) received another regimen. Complete remission was achieved in 70.4% (114/162) following initial treatment. CNS prophylaxis was administered in 33.3% (54/162): Intrathecal (IT) chemotherapy only in 16.7% (27/162), intravenous high‐dose methotrexate (HD‐MTX) only in 14.2% (23/162), and both in 2.5% (4/162).
At the time of CNS relapse, sites of involvement were the CNS parenchyma in 52.5% (85/162), the leptomeninges in 37.7% (61/162), and both compartments in 9.9% (16/162). Among patients with parenchymal involvement, the disease was multifocal in 65.4% (53/81). Parenchymal lesions were supratentorial (60%; 48/80), infratentorial (10%; 8/80), or both (30%; 24/80). Lesions greater than 1 cm were seen in 64.1% (50/78), and a midline shift was seen in 20.3% (15/74).
3.2. Therapeutic Intervention Characteristics and Treatment Responses in R‐SCNSL
Treatment regimens for 162 patients with CNS relapse were stratified by disease site into CNS‐only (n = 120) or concomitant CNS/systemic disease (n = 42) and categorized by intensity. Among patients with available and classified treatment data, the CNS‐only group received moderate‐intensity therapy in 44 patients (51.8%), low‐intensity in 35 (41.2%), and high‐intensity with MATRix in 6 (7.1%). In contrast, patients with concomitant CNS/systemic relapse more frequently received high‐intensity therapy, including R‐ICE (5 of 22 patients; 22.7%), compared with none in the CNS‐only group (Supporting Information S1: Table 1). Overall, the distribution of treatment intensity differed significantly between the two groups (p < 0.001). When high‐intensity regimens were analyzed as a combined category (MATRix/R‐ICE), the difference between groups was no longer statistically significant (p = 0.087) (Supporting Information S1: Table 1). Regarding consolidation, 24.7% (40/162) of the total cohort received consolidation therapy with autologous stem cell transplant (ASCT). The conditioning regimens included BEAM (carmustine [BCNU], etoposide, cytarabine, and melphalan) chemotherapy in 15.4% of patients (25/162), high‐dose BCNU plus thiotepa in 8.6% (14/162), and cyclophosphamide/total body irradiation (Cy/TBI) in 0.6% (1/162).
The overall response rate was 44.8%, with 38.2% achieving complete response (CR) and 6.6% achieving partial response (PR) (Supporting Information S1: Table 1). No significant difference in treatment response rates was observed between patients with CNS‐only relapse and those with concomitant CNS/systemic relapse (p = 0.24). In univariable Cox regression models, combined high‐intensity regimens (MATRix/R‐ICE) were significantly associated with inferior post‐relapse survival compared with moderate‐intensity therapy, with worse OS (HR 2.47, 95% CI 1.20–5.07; p = 0.014) and worse PFS (HR 2.41, 95% CI 1.21–4.81; p = 0.012) (Supporting Information S1: Tables 2 and 3). No significant differences in OS or PFS were observed when comparing low‐intensity to moderate‐intensity treatments. However, in multivariable Cox regression models for OS, high‐intensity therapy (MATRix/R‐ICE) remained independently associated with inferior OS compared with moderate‐intensity therapy (HR, 2.36; 95% CI, 1.08–5.14; p = 0.031), whereas low‐intensity therapy did not differ significantly from moderate‐intensity treatment (Table 1). However, given the small number of patients in the MATRix/R‐ICE group (n = 11), no definitive conclusion can be drawn.
TABLE 1.
Multivariable cox regression models predicting overall survival post‐CNS relapse in R‐SCNSL.
| Variables | Comparison | HR (95% CI) | p value |
|---|---|---|---|
| R‐SCNSL subtypes | Concomitant CNS/Systemic versus CNS‐only | 1.17 (0.66, 2.09) | 0.587 |
| Pathologic subtypes | DHL/THL versus DLBCL‐GCB | 1.48 (0.73, 3.02) | 0.280 |
| DLBCL‐non‐GCB versus DLBCL‐GCB | 1.06 (0.59, 1.9) | 0.851 | |
| HGBCL‐NOS versus DLBCL‐GCB | 1.22 (0.49, 3.03) | 0.662 | |
| ASCT post‐CNS relapse | ASCT versus No ASCT post‐CNS relapse | 0.33 (0.18, 0.58) | < 0.001 |
| Relapse CNS compartment | Leptomeningeal versus Parenchymal | 1.98 (1.2, 3.26) | 0.008 |
| Treatment intensity | Low versus Moderate intensity | 1.14 (0.66, 1.98) | 0.643 |
| High (MATRIX/R‐ICE) versus Moderate intensity | 2.36 (1.08, 5.14) | 0.031 |
Abbreviations: ASCT, autologous stem cell transplant; CNS, central nervous system; DHL, double‐hit lymphoma; DLBCL, diffuse large B‐cell lymphoma; GCB, germinal center B‐cell‐like; HGBCL, high‐grade B‐cell lymphoma; LM, leptomeningeal; NA, not available; non‐GCB, non‐germinal center B‐cell‐like; NOS, not otherwise specified; OS, overall survival; p, parenchymal; R‐SCNSL, relapsed secondary central nervous system lymphoma; THL, triple‐hit lymphoma; w/o, without; y, years.
Across the entire cohort, treatment data were unavailable for six patients (3.7%), either because treatment information was missing or because the patients died before treatment could be administered.
3.3. Clinical and Biological Variables Associated With CNS Anatomical Localization in R‐SCNSL
An analysis of associations between clinical and biological variables and the CNS anatomical compartment of relapse (leptomeningeal vs. parenchymal) was performed (Supporting Information S1: Table 1; Table 2). Leptomeningeal involvement was associated with a significantly higher prevalence of the DHL/THL pathologic status (p = 0.003), MYC gene rearrangement (p = 0.005), and absence of BCL6 expression (defined as < 30% positive staining lymphoma cells by IHC) (p = 0.005). Clinical factors associated with leptomeningeal disease included bone marrow involvement (p < 0.001), shorter time to relapse from initial diagnosis (p = 0.008), and concomitant CNS/systemic R‐SCNSL (p = 0.037; Table 2).
TABLE 2.
Association between the clinical/biological factors and relapse CNS compartment in R‐SCNSL.
| Parenchymal (N = 85) | Leptomeningeal (N = 61) | Total (N = 146) | p value | |
|---|---|---|---|---|
| Pathologic subtypes | 0.003 | |||
| DHL/THL | 14 (16.5%) | 24 (39.3%) | 38 (26.0%) | |
| DLBCL‐non‐GCB | 40 (47.1%) | 20 (32.8%) | 60 (41.1%) | |
| DLBCL‐GCB | 28 (32.9%) | 11 (18.0%) | 39 (26.7%) | |
| HGBCL‐NOS | 3 (3.5%) | 6 (9.8%) | 9 (6.2%) | |
| Cell of origin signature by Hans algorithm | 0.66 | |||
| N‐Miss | 17 | 30 | 47 | |
| GCB | 28 (41.2%) | 11 (35.5%) | 39 (39.4%) | |
| non‐GCB | 40 (58.8%) | 20 (64.5%) | 60 (60.6%) | |
| MYC rearrangement | 0.005 | |||
| N‐Miss | 30 | 12 | 42 | |
| Negative | 40 (72.7%) | 22 (44.9%) | 62 (59.6%) | |
| Positive | 15 (27.3%) | 27 (55.1%) | 42 (40.4%) | |
| BCL2 rearrangement | 0.053 | |||
| N‐Miss | 39 | 14 | 53 | |
| Negative | 34 (73.9%) | 25 (53.2%) | 59 (63.4%) | |
| Positive | 12 (26.1%) | 22 (46.8%) | 34 (36.6%) | |
| BCL6 rearrangement | 0.52 | |||
| N‐Miss | 43 | 17 | 60 | |
| Negative | 26 (61.9%) | 24 (54.5%) | 50 (58.1%) | |
| Positive | 16 (38.1%) | 20 (45.5%) | 36 (41.9%) | |
| BCL6 expression | 0.005 | |||
| N‐Miss | 7 | 7 | 14 | |
| Negative | 13 (16.7%) | 21 (38.9%) | 34 (25.8%) | |
| Positive | 65 (83.3%) | 33 (61.1%) | 98 (74.2%) | |
| MYD88 L265P | 0.32 | |||
| N‐Miss | 65 | 56 | 121 | |
| Negative | 9 (45.0%) | 4 (80.0%) | 13 (52.0%) | |
| Positive | 11 (55.0%) | 1 (20.0%) | 12 (48.0%) | |
| Age at CNS relapse | 0.45 | |||
| Median (Range) | 64.0 (23.0, 88.0) | 65.0 (24.0, 83.0) | 65.0 (23.0, 88.0) | |
| Mean (SD) | 62.2 (13.7) | 64.3 (11.7) | 63.1 (12.9) | |
| Race | 0.61 | |||
| Caucasian | 73 (85.9%) | 55 (90.2%) | 128 (87.7%) | |
| Non‐caucasian | 12 (14.1%) | 6 (9.8%) | 18 (12.3%) | |
| Sex | 0.13 | |||
| Female | 31 (36.5%) | 30 (49.2%) | 61 (41.8%) | |
| Male | 54 (63.5%) | 31 (50.8%) | 85 (58.2%) | |
| CNS‐IPI Score | 0.098 | |||
| N‐Miss | 32 | 16 | 48 | |
| Low | 10 (18.9%) | 2 (4.4%) | 12 (12.2%) | |
| Intermediate | 32 (60.4%) | 32 (71.1%) | 64 (65.3%) | |
| High | 11 (20.8%) | 11 (24.4%) | 22 (22.4%) | |
| Kidney/Adrenal involvement | 1.00 | |||
| No | 78 (91.8%) | 56 (91.8%) | 134 (91.8%) | |
| Yes | 7 (8.2%) | 5 (8.2%) | 12 (8.2%) | |
| Bone marrow involvement | < 0.001 | |||
| N‐Miss | 2 | 0 | 2 | |
| No | 74 (89.2%) | 38 (62.3%) | 112 (77.8%) | |
| Yes | 9 (10.8%) | 23 (37.7%) | 32 (22.2%) | |
| CNS Prophylaxis | 0.15 | |||
| Not received | 62 (72.9%) | 37 (60.7%) | 99 (67.8%) | |
| Received | 23 (27.1%) | 24 (39.3%) | 47 (32.2%) | |
| CNS Prophylaxis (IV HD‐MTX) | 0.21 | |||
| N‐Miss | 1 | 1 | 2 | |
| Not received | 76 (90.5%) | 50 (83.3%) | 126 (87.5%) | |
| Received | 8 (9.5%) | 10 (16.7%) | 18 (12.5%) | |
| CNS Prophylaxis (IT chemotherapy) | 0.52 | |||
| N‐Miss | 1 | 1 | 2 | |
| Not received | 70 (83.3%) | 47 (78.3%) | 117 (81.2%) | |
| Received | 14 (16.7%) | 13 (21.7%) | 27 (18.8%) | |
|
Time to relapse from initial lymphoma diagnosis (months) |
||||
| Median (Range) | 10.2 (1.5, 166.0) | 8.8 (0.9, 34.8) | 9.2 (0.9, 166.0) | 0.008 |
| Mean (SD) | 24.1 (30.5) | 10.2 (7.1) | 18.3 (24.7) | |
| R‐SCNSL subtypes | ||||
| CNS‐Only | 68 (80.0%) | 39 (63.9%) | 107 (73.3%) | 0.037 |
| Concomitant CNS/Systemic | 17 (20.0%) | 22 (36.1%) | 39 (26.7%) |
Abbreviations: BCL2, B‐cell lymphoma 2; BCL6, B‐cell lymphoma 6; CNS, central nervous system; CNS‐IPI, Central Nervous System International Prognostic Index; DHL, double‐hit lymphoma; DLBCL, diffuse large B‐cell lymphoma; GCB, germinal center B‐cell‐like; HGBCL, high‐grade B‐cell lymphoma; IT, intrathecal; IV HD‐MTX, intravenous high‐dose methotrexate; MYC, myelocytomatosis oncogene; MYD88 L265P, myeloid differentiation primary response 88 L265P mutation; N‐Miss, number missing; non‐GCB, non‐germinal center B‐cell‐like; NOS, not otherwise specified; R‐SCNSL, relapsed secondary central nervous system lymphoma; SD, standard deviation; THL, triple‐hit lymphoma.
In contrast, other variables were not associated with preferential CNS compartment involvement, including: cell of origin by Hans algorithm, BCL2 or BCL6 rearrangement, MYD88 L265P mutation status, age at relapse, race, sex, CNS‐IPI score, kidney/adrenal involvement, and CNS prophylaxis with intrathecal or intravenous chemotherapy.
3.4. Clinical and Biological Variables Associated With OS and PFS in R‐SCNSL
After a median follow‐up of 8 months (range: 0 days to 173 months), a total of 113 deaths occurred. Based on reversed Kaplan–Meier methods, the follow‐up rates were 88.1%, 81.7% and 62.5% at 1, 3 and 5 years, respectively. OS rates for the whole cohort at 1, 2, and 5 years after CNS relapse were 47.6% (95% CI: 40.28%–56.33%), 36.87% (95% CI: 29.82%–45.59%), and 27.67% (95% CI: 21.16%–36.19%), respectively (Table 3). Univariable Cox regression analysis showed that SCNSL classification (CNS‐only vs. concomitant), pathologic subtype, relapse CNS compartment (leptomeningeal vs. parenchymal) and receipt of ASCT after CNS relapse were all associated with OS (Supporting Information S1: Table 2). Receipt of ASCT and relapse CNS compartment remained statistically significant in the multivariable model (Table 1). After adjusting for SCNSL classification, pathologic subtype, and CNS relapse compartment, patients who underwent ASCT had a 67% reduced hazard for death compared to those who did not (HR = 0.33, 95% CI: 0.18–0.58, p < 0.001). In contrast, leptomeningeal relapse was associated with a 98% increase in hazard for death compared to parenchymal relapse (HR = 1.98, 95% CI: 1.20–3.26, p = 0.008) (Table 1). Variables evaluated but not included in the final model are described in the Methods.
TABLE 3.
Kaplan‐Meier estimates of overall survival post‐CNS relapse in R‐SCNSL.
| Variables | Total | Total Events | 1‐year OS | 2‐year OS | 5‐year OS | p value |
|---|---|---|---|---|---|---|
| All | 162 | 113 | 47.63% (40.28%, 56.33%) | 36.87% (29.82%, 45.59%) | 27.67% (21.16%, 36.19%) | — |
| R‐SCNSL subtypes | 0.026 | |||||
| CNS‐Only | 120 | 80 | 52.61% (44.1%, 62.7%) | 43.02% (34.7%, 53.4%) | 32.73% (24.8%, 43.1%) | — |
| Concomitant CNS/Systemic | 42 | 33 | 32.89% (20.8%, 51.9%) | 18.50% (9.3%, 36.9%) | 12.33% (5.1%, 30.1%) | — |
| Pathologic subtypes | 0.010 | |||||
| DHL/THL | 43 | 30 | 33.13% (20.53%, 53.46%) | 26.50% (14.99%, 46.87%) | 22.72% (11.91%, 43.31%) | — |
| DLBCL‐non‐GCB | 66 | 45 | 52.57% (41.55%, 66.52%) | 40.81% (30.18%, 55.18%) | 31.52% (21.58%, 46.02%) | — |
| DLBCL‐GCB | 44 | 29 | 58.44% (45.44%, 75.16%) | 46.27% (33.44%, 64.01%) | 33.92% (22.18%, 51.87%) | — |
| HGBCL‐NOS | 9 | 9 | 22.22% (6.55%, 75.44%) | 11.11% (1.75%, 70.51%) | 0.00% (NA, NA) | — |
| MYC rearrangement | < 0.001 | |||||
| Negative | 71 | 41 | 63.3% (52.8%, 75.9%) | 43.5% (32.8%, 57.6%) | 40.0% (29.5%, 54.2%) | |
| Positive | 47 | 34 | 29.7% (18.1%, 48.6%) | 20.3% (10.5%, 39.2%) | 20.3% (10.5%, 39.2%) | |
| BCL6 expression | 0.005 | |||||
| Negative | 36 | 31 | 45.75% (31.90%, 65.63%) | 25.74% (14.66%, 45.19%) | 10.72% (3.99%, 28.83%) | |
| Positive | 54.18% (45.27%, 64.84%) | 45.40% (36.52%, 56.43%) | 37.24% (28.63%, 48.43%) | |||
| Bone marrow involvement | 0.001 | |||||
| No | 122 | 82 | 55.0% (46.6%, 64.920%) | 43.8% (35.5%, 54.051%) | 31.9% (24.1%, 42.099%) | |
| Yes | 38 | 31 | 23.2% (12.7%, 42.169%) | 13.5% (5.7%, 32.089%) | 13.5% (5.7%, 32.089%) | |
| Treatment intensity (R‐SCNSL) | 0.035 | |||||
| Moderate | 52 | 34 | 56.0% (43.8%, 71.711%) | 45.4% (33.3%, 61.795%) | 36.7% (25.3%, 53.292%) | — |
| Low | 44 | 33 | 54.1% (40.8%, 71.654%) | 41.8% (29.1%, 59.969%) | 25.2% (14.4%, 43.957%) | — |
| High (MATRIX) | 6 | 5 | 50.0% (22.5%, 100.000%) | 16.7% (2.8%, 99.744%) | 16.7% (2.8%, 99.744%) | — |
| High (R‐ICE) | 5 | 5 | 0.0% (NA, NA) | 0.0% (NA, NA) | 0.0% (NA, NA) | — |
| Treatment intensity (R‐SCNSL; combined high‐intensity regimens) | 0.042 | |||||
| Moderate | 52 | 34 | 56.0% (43.8%, 71.711%) | 45.4% (33.3%, 61.795%) | 36.7% (25.3%, 53.292%) | — |
| Low | 44 | 33 | 54.1% (40.8%, 71.654%) | 41.8% (29.1%, 59.969%) | 25.2% (14.4%, 43.957%) | — |
| High (MATRIX/R‐ICE) | 11 | 10 | 27.3% (10.4%, 71.587%) | 9.1% (1.4%, 58.911%) | 9.1% (1.4%, 58.911%) | — |
| ASCT post‐CNS relapse | < 0.001 | |||||
| ASCT | 40 | 21 | 75.0% (62.7%, 89.7%) | 67.2% (54.0%, 83.6%) | 56.0% (42.2%, 74.3%) | — |
| No ASCT | 121 | 91 | 38.2% (30.1%, 48.6%) | 26.2% (18.9%, 36.1%) | 17.3% (11.2%, 26.8%) | — |
| ASCT post‐CNS relapse (4 months landmark analysis) | 0.013 | |||||
| ASCT | 38 | 19 | 73.7% (60.9%, 89.1%) | 67.8% (54.3%, 84.7%) | 58.9% (44.9%, 77.4%) | — |
| No ASCT | 61 | 42 | 59.4% (48.1%, 73.4%) | 43.7% (32.6%, 58.5%) | 27.4% (17.5%, 42.8%) | |
| ASCT post‐CNS relapse (4‐month landmark analysis in CR/PR patients) | 0.96 | |||||
| ASCT | 35 | 17 | 74.3% (61.1%, 90.3%) | 67.8% (53.7%, 85.5%) | 61.4% (46.8%, 80.4%) | |
| No ASCT | 25 | 11 | 79% (65.4%, 97.3%) | 67% (50.9%, 88.7%) | 50.9% (33.2%, 77.96%) | |
| Relapse CNS compartment (Both compartment involvement under LM) | < 0.001 | |||||
| Leptomeningeal | 77 | 59 | 28.9% (19.9%, 42.0%) | 17.6% (10.3%, 29.9%) | 14.0% (7.6%, 26.0%) | — |
| Parenchymal | 85 | 54 | 63.5% (53.9%, 74.9%) | 53.0% (43.1%, 65.2%) | 39.4% (29.9%, 52.1%) | — |
| Relapse CNS compartment | < 0.001 | |||||
| Both compartments | 16 | 11 | 40.38% (21.82%, 74.75%) | 33.65% (16.51%, 68.60%) | 25.24% (10.16%, 62.67%) | — |
| Leptomeningeal | 61 | 48 | 25.66% (16.19%, 40.67%) | 13.16% (6.42%, 26.96%) | 10.97% (4.92%, 24.44%) | — |
| Parenchymal | 85 | 54 | 63.50% (53.87%, 74.85%) | 53.04% (43.15%, 65.21%) | 39.45% (29.89%, 52.06%) | — |
| Anatomical classification of R‐SCNSL | < 0.001 | |||||
| LM‐CNS‐only | 52 | 38 | 31.43% (20.53%, 48.12%) | 22.04% (12.69%, 38.28%) | 19.59% (10.77%, 35.64%) | — |
| LM‐concomitant | 25 | 21 | 23.72% (11.11%, 50.65%) | 7.12% (1.26%, 40.32%) | 0.00% (NA, NA) | — |
| P‐CNS‐Only | 68 | 42 | 68.02% (57.60%, 80.33%) | 58.22% (47.29%, 71.67%) | 42.75% (31.95%, 57.21%) | — |
| P‐concomitant | 17 | 12 | 45.29% (26.49%, 77.46%) | 32.35% (15.87%, 65.96%) | 25.88% (11.21%, 59.73%) | — |
| Time of relapse and relapse CNS compartment | < 0.001 | |||||
| Early‐LM | 55 | 42 | 26.66% (16.57%, 42.91%) | 15.55% (7.92%, 30.53%) | 13.33% (6.37%, 27.92%) | — |
| Early‐P | 46 | 31 | 56.98% (44.06%, 73.70%) | 47.49% (34.71%, 64.96%) | 40.07% (27.73%, 57.90%) | — |
| Late‐LM | 22 | 17 | 34.09% (18.77%, 61.90%) | 21.92% (9.13%, 52.63%) | 14.61% (4.46%, 47.86%) | — |
| Late‐P | 39 | 23 | 71.19% (58.15%, 87.16%) | 59.54% (45.55%, 77.83%) | 38.70% (25.46%, 58.82%) | — |
| ASCT post‐CNS relapse and relapse CNS compartment | < 0.001 | |||||
| LM‐W/O ASCT | 64 | 52 | 19.4% (11.26%, 33.58%) | 9.7% (4.26%, 22.21%) | 7.8% (3.05%, 19.82%) | — |
| LM‐with ASCT | 13 | 7 | 69.2% (48.19%, 99.47%) | 51.3% (29.32%, 89.71%) | 41.0% (20.16%, 83.49%) | — |
| P‐W/O ASCT | 58 | 40 | 56.6% (44.92%, 71.44%) | 42.5% (30.99%, 58.25%) | 27.4% (17.27%, 43.35%) | — |
| P‐with ASCT | 27 | 14 | 77.8% (63.58%, 95.15%) | 74.1% (59.26%, 92.59%) | 62.4% (46.39%, 83.88%) | — |
Abbreviations: ASCT, autologous stem cell transplant; CNS, central nervous system; DHL, double‐hit lymphoma; DLBCL, diffuse large B‐cell lymphoma; GCB, germinal center B‐cell‐like; HGBCL, high‐grade B‐cell lymphoma; LM, leptomeningeal; NA, not available; non‐GCB, non‐germinal center B‐cell‐like; NOS, not otherwise specified; OS, overall survival; p, parenchymal; R‐SCNSL, relapsed secondary central nervous system lymphoma; THL, triple‐hit lymphoma; w/o, without; y, years.
PFS rates for the whole cohort at 1, 2, and 5 years were 28.2% (95% CI: 21.93%–36.29%), 18.6% (95% CI: 13.34%–26.01%), and 14.4% (95% CI: 9.68%–21.29%), respectively (Table 4). SCNSL classification (CNS‐only vs. concomitant), pathologic subtype, receipt of ASCT after CNS relapse, and relapse CNS compartment (leptomeningeal vs. parenchymal) were all significantly associated with PFS in univariable analyses (Supporting Information S1: Table 3). In multivariable models, receipt of ASCT remained independently associated with improved PFS, while leptomeningeal relapse showed a trend toward inferior PFS that did not reach statistical significance (Table 5).
TABLE 4.
Kaplan‐Meier estimates of progression free survival post‐CNS relapse in R‐SCNSL.
| Variables | Total | Total events | 1‐year PFS | 2‐year PFS | 5‐year PFS | p value |
|---|---|---|---|---|---|---|
| All | 162 | 136 | 28.21% (21.93%, 36.29%) | 18.62% (13.34%, 26.01%) | 14.35% (9.68%, 21.29%) | — |
| R‐SCNSL subtypes | 0.010 | |||||
| CNS‐only | 120 | 99 | 32.6% (25.09%, 42.46%) | 22.4% (15.88%, 31.66%) | 16.6% (10.89%, 25.32%) | — |
| Concomitant CNS/Systemic | 42 | 37 | 15.3% (7.34%, 31.87%) | 7.6% (2.58%, 22.65%) | 7.6% (2.58%, 22.65%) | — |
| Pathologic subtypes | 0.003 | |||||
| DHL/THL | 43 | 37 | 10.81% (4.30%, 27.1%) | 7.20% (2.13%, 24.4%) | 7.20% (2.13%, 24.4%) | — |
| DLBCL‐non‐GCB | 66 | 54 | 35.61% (25.53%, 49.7%) | 21.04% (13.00%, 34.0%) | 16.02% (9.04%, 28.4%) | — |
| DLBCL‐GCB | 44 | 36 | 36.36% (24.60%, 53.8%) | 29.09% (18.26%, 46.3%) | 21.82% (12.35%, 38.5%) | — |
| HGBCL‐NOS | 9 | 9 | 11.11% (1.75%, 70.5%) | 0.00% (NA, NA) | 0.00% (NA, NA) | — |
| MYC rearrangement | < 0.001 | |||||
| Negative | 71 | 54 | 40.4% (30.2%, 53.9%) | 28.0% (19.0%, 41.1%) | 21.6% (13.6%, 34.3%) | — |
| Positive | 47 | 41 | 9.8% (3.9%, 24.7%) | 6.5% (1.9%, 22.1%) | 6.5% (1.9%, 22.1%) | — |
| BCL6 expression | 0.23 | |||||
| Negative | 36 | 33 | 25.7% (14.7%, 45.19%) | 11.4% (4.5%, 28.76%) | 5.7% (1.5%, 21.96%) | — |
| Positive | 109 | 88 | 32.5% (24.6%, 42.87%) | 23.1% (16.2%, 32.99%) | 18.9% (12.6%, 28.40%) | — |
| Bone marrow involvement | 0.014 | |||||
| No | 122 | 101 | 32.18% (24.72%, 41.9%) | 21.22% (14.88%, 30.3%) | 15.47% (9.99%, 24.0%) | — |
| Yes | 38 | 34 | 16.24% (7.81%, 33.8%) | 10.83% (4.29%, 27.3%) | 10.83% (4.29%, 27.3%) | — |
| Treatment intensity (R‐SCNSL) | 0.066 | |||||
| Moderate | 52 | 41 | 35.81% (24.79%, 51.7%) | 29.59% (19.32%, 45.3%) | 25.36% (15.74%, 40.9%) | — |
| Low | 44 | 38 | 29.69% (18.51%, 47.6%) | 14.85% (7.10%, 31.0%) | 9.28% (3.43%, 25.1%) | — |
| High (MATRIX) | 6 | 6 | 0.00% (NA, NA) | 0.00% (NA, NA) | 0.00% (NA, NA) | — |
| High (R‐ICE) | 5 | 5 | 0.00% (NA, NA) | 0.00% (NA, NA) | 0.00% (NA, NA) | — |
| Treatment intensity (R‐SCNSL; combined high‐intensity regimens) | 0.035 | |||||
| Moderate | 52 | 41 | 35.81% (24.79%, 51.7%) | 29.59% (19.32%, 45.3%) | 25.36% (15.74%, 40.9%) | — |
| Low | 44 | 38 | 29.69% (18.51%, 47.6%) | 14.85% (7.10%, 31.0%) | 9.28% (3.43%, 25.1%) | — |
| High (MATRIX/R‐ICE) | 11 | 11 | 0.00% (NA, NA) | 0.00% (NA, NA) | 0.00% (NA, NA) | — |
| ASCT post‐CNS relapse | < 0.001 | |||||
| ASCT | 40 | 29 | 55.0% (41.56%, 72.8%) | 39.4% (26.70%, 58.1%) | 34.1% (22.06%, 52.8%) | — |
| No ASCT | 121 | 106 | 19.0% (13.00%, 27.9%) | 11.4% (6.77%, 19.3%) | 7.3% (3.70%, 14.6%) | — |
| ASCT post‐CNS relapse (4‐month landmark analysis) | 0.004 | |||||
| ASCT | 38 | 27 | 57.9% (44.14%, 75.9%) | 41.4% (28.29%, 60.7%) | 35.9% (23.36%, 55.2%) | — |
| No ASCT | 62 | 53 | 34.4% (24.35%, 48.7%) | 20.7% (12.55%, 34.0%) | 13.3% (6.81%, 25.9%) | |
| ASCT post‐CNS relapse (4‐month landmark analysis in CR/PR patients) | 0.36 | |||||
| ASCT | 35 | 24 | 60% (45.8%, 78.6%) | 45% (31.2%, 65.3%) | 39% (25.7%, 59.5%) | |
| No ASCT | 25 | 18 | 52% (35.7%, 75.8%) | 39% (23.7%, 64.2%) | 25% (12.5%, 51.2%) | |
| Relapse CNS compartment | < 0.001 | |||||
| Leptomeningeal | 77 | 67 | 16.8% (10.0%, 28.1%) | 9.4% (4.6%, 19.6%) | 7.9% (3.5%, 17.7%) | — |
| Parenchymal | 85 | 69 | 38.3% (29.1%, 50.4%) | 26.8% (18.7%, 38.5%) | 20.3% (13.1%, 31.4%) | — |
| Relapse CNS compartment | < 0.001 | |||||
| Both compartments | 16 | 12 | 37.50% (19.92%, 70.6%) | 23.44% (9.33%, 58.9%) | 23.44% (9.33%, 58.9%) | — |
| Leptomeningeal | 61 | 55 | 10.81% (5.09%, 23.0%) | 5.41% (1.80%, 16.2%) | 3.60% (0.93%, 14.0%) | — |
| Parenchymal | 85 | 69 | 38.29% (29.09%, 50.4%) | 26.80% (18.68%, 38.5%) | 20.27% (13.09%, 31.4%) | — |
| Anatomical classification of R‐SCNSL | < 0.001 | |||||
| LM‐CNS‐only | 52 | 44 | 20.30% (11.69%, 35.247%) | 13.92% (6.90%, 28.069%) | 11.60% (5.28%, 25.492%) | — |
| LM‐concomitant | 25 | 23 | 9.01% (2.41%, 33.666%) | 0.00% (NA, NA) | 0.00% (NA, NA) | — |
| P‐CNS‐only | 68 | 55 | 42.00% (31.56%, 55.893%) | 29.08% (19.81%, 42.682%) | 20.77% (12.78%, 33.751%) | — |
| P‐concomitant | 17 | 14 | 23.53% (9.99%, 55.436%) | 17.65% (6.32%, 49.277%) | 17.65% (6.32%, 49.277%) | — |
| Time of relapse and relapse CNS compartment | 0.004 | |||||
| Early‐LM | 55 | 47 | 12.149% (5.75%, 25.65%) | 8.100% (3.17%, 20.68%) | 8.100% (3.17%, 20.68%) | — |
| Early‐P | 46 | 38 | 35.556% (23.99%, 52.69%) | 26.667% (16.43%, 43.29%) | 22.000% (12.65%, 38.27%) | — |
| Late‐LM | 22 | 20 | 27.273% (13.78%, 53.96%) | 12.121% (3.67%, 40.01%) | 6.061% (0.97%, 37.76%) | — |
| Late‐P | 39 | 31 | 41.261% (28.08%, 60.63%) | 26.525% (15.33%, 45.89%) | 17.683% (8.63%, 36.22%) | — |
| ASCT post‐CNS relapse and relapse CNS compartment | < 0.001 | |||||
| LM‐W/O ASCT | 64 | 58 | 10.34% (4.86%, 22.0%) | 5.17% (1.72%, 15.5%) | 3.45% (0.88%, 13.4%) | — |
| LM‐with ASCT | 13 | 9 | 46.15% (25.66%, 83.0%) | 28.85% (11.84%, 70.3%) | 28.85% (11.84%, 70.3%) | — |
| P‐W/O ASCT | 58 | 49 | 28.02% (18.31%, 42.9%) | 18.02% (10.11%, 32.1%) | 11.68% (5.44%, 25.1%) | — |
| P‐with ASCT | 27 | 20 | 59.26% (43.34%, 81.0%) | 44.44% (29.15%, 67.8%) | 37.04% (22.65%, 60.6%) | — |
Abbreviations: ASCT, autologous stem cell transplant; CNS, central nervous system; DHL, double‐hit lymphoma; DLBCL, diffuse large B‐cell lymphoma; GCB, germinal center B‐cell‐like; HGBCL, high‐grade B‐cell lymphoma; LM, leptomeningeal; NA, not available; non‐GCB, non‐germinal center B‐cell‐like; NOS, not otherwise specified; OS, overall survival; p, parenchymal; R‐SCNSL, relapsed secondary central nervous system lymphoma; THL, triple‐hit lymphoma; w/o, without; y, years.
TABLE 5.
Multivariable cox regression models predicting progression‐free survival post‐CNS relapse in R‐SCNSL.
| Variables | Comparison | HR (95% CI) | p value |
|---|---|---|---|
| R‐SCNSL subtypes | Concomitant CNS/Systemic versus CNS‐only | 1.4 (0.81, 2.43) | 0.23 |
| Pathologic subtypes | DHL/THL versus DLBCL‐GCB | 1.53 (0.82, 2.84) | 0.18 |
| DLBCL‐non‐GCB versus DLBCL‐GCB | 0.82 (0.48, 1.38) | 0.45 | |
| HGBCL‐NOS versus DLBCL‐GCB | 0.95 (0.39, 2.31) | 0.91 | |
| ASCT post‐CNS relapse | ASCT versus No ASCT post‐CNS relapse | 0.36 (0.22, 0.61) | < 0.001 |
| Relapse CNS compartment | Leptomeningeal versus Parenchymal | 1.52 (0.97, 2.4) | 0.07 |
| Treatment intensity | Low versus Moderate intensity | 1.16 (0.71, 1.9) | 0.55 |
| High (MATRIX/R‐ICE) versus Moderate | 2.08 (0.97, 4.45) | 0.06 |
Abbreviations: ASCT, autologous stem cell transplant; CNS, central nervous system; DHL, double‐hit lymphoma; DLBCL, diffuse large B‐cell lymphoma; GCB, germinal center B‐cell‐like; HGBCL, high‐grade B‐cell lymphoma; LM, leptomeningeal; NA, not available; non‐GCB, non‐germinal center B‐cell‐like; NOS, not otherwise specified; OS, overall survival; p, parenchymal; R‐SCNSL, relapsed secondary central nervous system lymphoma; THL, triple‐hit lymphoma; w/o, without; y, years.
3.5. Survival Analysis by Anatomical Classification of R‐SCNSL
Based on three anatomic localizations (leptomeningeal, parenchymal, and systemic), we classified R‐SCNSL into the following four anatomical subtypes: (1) CNS‐only R‐SCNSL with CNS parenchymal‐only involvement (P‐CNS): 68 patients (42%); (2) Concomitant R‐SCNSL with CNS parenchymal‐only involvement (P‐concomitant): 17 patients (10.5%); (3) CNS‐only R‐SCNSL with leptomeningeal involvement, with or without CNS parenchymal involvement (LM‐CNS): 52 patients (32.1%); and (4) Concomitant R‐SCNSL with leptomeningeal involvement, with or without CNS parenchymal involvement (LM‐concomitant): 25 patients (15.4%).
By Kaplan–Meier analysis, the anatomical classification was associated with significant impact on OS and PFS (p < 0.001) (Figures 1 and 2). 1‐year, 2‐year, and 5‐year OS are 68%, 58.2%, and 42.8% (P‐CNS); 45.3%, 32.4%, and 25.9% (P‐concomitant); 31.4%, 22%, and 19.6% (LM‐CNS); and 23.7%, 7.1%, and 0% (LM‐concomitant) (Table 3). 1‐year, 2‐year, and 5‐year PFS are 42%, 29.1%, and 20.8% (P‐CNS); 23.5%, 17.7%, and 17.7% (P‐concomitant); 20.3%, 13.9%, and 11.6% (LM‐CNS); and 9%, 0%, and 0% (LM‐concomitant) (Table 4).
FIGURE 1.
Kaplan‐Meier overall survival curves by anatomical classification of R‐SCNSL.
FIGURE 2.
Kaplan‐Meier progression free survival curves by anatomical classification of R‐SCNSL.
3.6. Survival Impact of ASCT in R‐SCNSL
Multivariable analyses of OS and PFS showed that ASCT had a significant impact on both outcomes (Tables 1 and 5; both p < 0.001). At baseline, the 3‐year OS was 56% for the ASCT group versus 17.3% for the non‐ASCT group (Table 3), while the 3‐year PFS was 34.1% versus 7.3%, respectively (Table 4).
To address potential bias, we also performed a 4‐month landmark analysis and a 4‐month landmark analysis incorporating response status (CR/PR). In the 4‐month landmark analysis, ASCT remained significantly associated with improved OS (HR 0.50, 95% CI 0.29–0.87; p = 0.015; Supporting Information S1: Table 2) and PFS (HR 0.50, 95% CI 0.31–0.81; p = 0.0045; Supporting Information S1: Table 3).
When response status (CR/PR) was included in the landmark analysis, statistical significance was no longer observed for OS (HR 1.02, 95% CI 0.47–2.19; p = 0.96) or PFS (HR 0.75, 95% CI 0.40–1.40; p = 0.37), although a trend toward improved outcomes was observed. In this response‐adjusted subset, the 3‐year OS was 61.4% for the ASCT group versus 50.9% for the non‐ASCT group (Table 3), and the 3‐year PFS was 39% versus 25% (Table 4).
The number of patients eligible for these analyses decreased notably in the non‐ASCT group, from 121 patients at baseline to 62 in the 4‐month landmark analysis and to 25 in the 4‐month landmark plus response analysis. In contrast, the ASCT group showed only a small decrease, from 40 patients at baseline to 38 and 35 patients in the respective analyses.
4. Discussion
This study evaluates clinical and biological factors significantly associated with preferential involvement of CNS compartments and proposes an anatomical classification for R‐SCNSL. CNS‐only R‐SCNSL was more frequent than concomitant R‐SCNSL in our cohort, consistent with the largest international study of R‐SCNSL. [8] In terms of CNS compartment involvement, the findings are also similar, with parenchymal involvement being more common than leptomeningeal involvement. [8] It has been reported in the literature that the pattern of CNS anatomical localization in SCNSL has shifted over time. Historically, leptomeningeal involvement was predominant accounting for two‐thirds of cases while brain parenchymal involvement represented about one third of cases. [13] More recent series indicate that with advanced MRI techniques, parenchymal involvement is more frequently observed. For instance, one MRI‐based study indicated 85.7% parenchymal involvement with only a minority of cases showing isolated leptomeningeal involvement. [14, 15] Notably, while parenchymal involvement remained the most common pattern in the overall R‐SCNSL cohort (52.5%) and in CNS‐only R‐SCNSL (56.7%), leptomeningeal involvement was more frequent (52.4% vs. 40.5%) in the concomitant R‐SCNSL group. Our data also show that the CNS localization of relapse differs by timing. Early relapses (≤ 1 year) were more frequently leptomeningeal (54.9%), while late relapses (> 1 year) were predominantly parenchymal (65.0%). These findings suggest that CNS compartment‐specific involvement in SCNSL may reflect underlying biological heterogeneity rather than purely anatomic variation.
In terms of CNS parenchymal involvement, our R‐SCNSL cohort demonstrated a markedly more aggressive and disseminated phenotype compared with primary CNS lymphoma (PCNSL), which typically presents as a solitary supratentorial, and periventricular mass. [16] Instead of a single lesion, 65.4% of patients had multifocal parenchymal deposits, and nearly one‐third (30.0%) exhibited disease spanning both supra‐ and infratentorial compartments. When measurable, 64.1% of lesions exceeded 1 cm in diameter, and one in five cases (20.3%) showed a midline shift. Thus, parenchymal disease in R‐SCNSL appears to be more widespread and burdensome than in PCNSL. However, no specific characteristics of parenchymal disease were found to have a significant impact on survival.
Leptomeningeal involvement, in contrast, has been associated with inferior survival in this study and prior reports. [3, 8, 17, 18] Leptomeningeal involvement was more common than parenchymal involvement in our concomitant R‐SCNSL group, consistent with previous studies [3, 19], and likely further contributing to poorer survival. One explanation is that parenchymal involvement tends to be localized, whereas leptomeningeal (CSF pathway) involvement is often diffuse and can involve the whole neural axis. Moreover, effective drug delivery throughout the leptomeningeal compartment in CNS lymphoma is challenging due to CSF flow obstruction and the blood–CSF barrier. Uniform distribution of intrathecally administered drugs in the leptomeningeal compartment cannot be achieved in the presence of tumor masses. CSF flow obstruction can lead to uneven drug distribution, with suboptimal concentrations in some areas and elevated concentrations in others, increasing the risk of toxicity. [20, 21, 22] As intrathecal drugs typically penetrate only 1–2 mm [23], adequate tumor penetration can be a major issue. Furthermore, intrathecal administration can be associated with epidural and subdural leakage. [24] A CSF flow scan has been recommended before intrathecal therapy if flow blockage is suspected. [25, 26, 27] If flow obstruction is found, local measures such as surgery or focal radiation have been recommended before intrathecal chemotherapy. [25, 26, 27] With systemic administration, the blood–CSF barrier presents an obstacle to achieving therapeutic concentrations at the target site. [28] The diffuse nature of leptomeningeal disease further complicates uniform drug distribution throughout all CSF compartments. [21, 28].
Our study identified several clinical and biological variables that were significantly associated with CNS compartment involvement. DHL/THL pathologic status MYC gene rearrangement, negative BCL6 expression by IHC, bone marrow involvement, and concomitant/systemic R‐SCNSL were significantly associated with a higher likelihood of leptomeningeal involvement. Some of these findings are consistent with prior reports. [8] While all patients with positive MYD88 L265P mutation (13/13) belonged to the CNS‐only R‐SCNSL (p = 0.016), the mutation did not statistically predict a specific CNS compartment, likely due to the small sample size. We acknowledge the biopsy source, as 9 of the 13 positive cases were identified from CNS tissue. However, there was no clear evidence to suggest that tissue type (CNS vs. non‐CNS) alone accounted for the finding, which is more likely reflective of underlying disease biology. The presence of the mutation in systemic DLBCL has been reported as a strong predictor of CNS relapse with cumulative CNS relapse rates as high as 18.8% at 6 years, suggesting that MYD88 L265P mutation is important for CNS tropism and dissemination. [19, 29] These biomarkers can potentially assist with CNS evaluations and guide CNS‐targeted therapy. Other variables, including cell of origin signature by Hans algorithm, BCL2 rearrangement, BCL6 rearrangement, age at relapse, race, sex, CNS‐IPI score, kidney/adrenal involvement, and CNS prophylaxis with IT or IV chemotherapy, were not associated with preferential CNS compartment involvement.
The CNS‐IPI score was found to have a very low positive predictive value in our R‐SCNSL cohort, with only 21.9% of patients classified in the high‐risk group. This may have contributed to only 33.3% of patients receiving CNS prophylaxis. Our findings are in line with reports indicating that CNS‐IPI has a high negative predictive value but only modest positive predictive value. [11, 30, 31, 32] The limitation of the CNS‐IPI in predicting CNS relapse is likely related to the lack of incorporation of biological and genetic factors such as cell of origin signature and double‐hit/triple‐hit signature as well as high‐risk extranodal anatomical sites such as testicular, breast, or uterine involvement. [6, 12, 30, 33] Moreover, CNS‐IPI cannot discern between leptomeningeal and parenchymal relapse. [32] Clearly, a better predictive tool is needed for CNS risk assessment.
As for the survival outcomes, our cohort represents a relatively younger population. Considering that the median age at diagnosis for DLBCL is in the mid‐60s [12], our cohort, with a median age of 63 years, represents a relatively younger population of patients with aggressive B‐cell lymphoma. However, the median age at SCNSL diagnosis in our cohort was 65 years, which is at the upper limit of the reported median age range of 59–64 years for SCNSL. [2, 3, 34] As younger patients typically have fewer comorbidities and are eligible for intensive therapeutic interventions, this may have a positive impact on survival outcomes. [12] In addition to clinical factors, BCL6 expression may be prognostically relevant because it reflects multiple mechanisms of BCL6 deregulation, whereas BCL6 rearrangement represents only one pathway. [35] Accordingly, expression therefore captures a broader biologic subset, which may explain why it correlated with outcomes in our cohort while rearrangements did not. [36] In the univariable analysis, there was no significant association found between post‐CNS relapse survival and age (> 60 vs. ≤ 60 years), sex, race (Caucasian vs. non‐Caucasian), or timing of relapse (early [≤ 1 year] vs. late [> 1 year]). This was true for both overall survival (OS) (Table 3) and progression‐free survival (PFS) (Table 4).
High‐intensity therapeutic interventions (MATRix/RICE) were associated with inferior survival compared with the intermediate‐intensity therapeutic intervention group in our cohort. However, this comparison is likely underpowered due to the very small number of patients in the MATRix/RICE group. In addition, potential confounding factors, including differences in disease burden within this retrospective cohort, may have influenced the results. As such, no definitive conclusions can be drawn, and these findings should not guide treatment selection in routine clinical practice.
With respect to anatomical compartment, LM involvement was significantly associated with inferior OS and showed a trend toward inferior PFS. LM involvement was a poor prognostic indicator across all subgroups, including CNS‐only versus concomitant disease, early versus late relapse, and across pathologic subtypes. Patients with isolated leptomeningeal relapse experienced significantly shorter CNS‐related survival than those with parenchymal‐only disease, and outcomes among the mixed LM + P group were similarly poor.
We performed a comprehensive analysis of the survival impact of ASCT in R‐SCNSL, including objective assessment of biases, such as sustained sensitivity to treatment and response status (CR/PR). Without accounting for these two biases, ASCT was associated with a significant improvement in OS and PFS, which was sustained in the 4‐month landmark analysis. However, when both biases were taken into consideration, no significant differences were observed. These findings indicate that selection biases need to be considered, particularly response status, as the non‐ASCT comparison group had more aggressive and treatment‐resistant disease. Overall, our findings suggest that patients with treatment‐sensitive R‐SCNSL showed a trend toward a survival benefit from consolidative ASCT. Consistent with this, four phase II clinical trials for SCNSL have incorporated intensified CNS‐penetrating chemoimmunotherapy followed by ASCT and have demonstrated improved survival outcomes. [4, 5, 37, 38].
Based on the localization in three anatomic sites (leptomeningeal, parenchymal, and systemic), we have formulated an anatomical classification system for R‐SCNSL consisting of four groups: P‐CNS, P‐concomitant, LM‐CNS, and LM‐concomitant. LM groups include mixed (LM + P) cases, in addition to LM‐alone cases, as LM involvement denotes diffuse neural axis involvement and poor prognosis. Parenchymal groups have isolated parenchymal involvement. This classification has high clinical utility, as it predicts prognosis and survival with a clear delineation of survival curves between groups. The poorest survival is observed in the LM‐concomitant group, whereas the best survival is observed in the P‐CNS group. Moreover, the classification can guide management. For concomitant groups, systemic disease‐targeted therapy should be considered along with CNS‐targeted therapy. Classification of SCNSL as leptomeningeal or parenchymal was determined by available MRI scans and CSF analysis. Some patients (27/162, 16.7%) did not undergo CSF studies; however, MRI alone does not completely rule out leptomeningeal involvement. As such, the accuracy of the classification is limited by the absence of CSF studies.
In conclusion, we propose that our anatomical localization‐based classification of R‐SCNSL into four groups (P‐CNS, P‐concomitant, LM‐CNS, and LM‐concomitant) is clinically useful, as it predicts prognosis and can guide therapeutic interventions. Biomarkers identified in this study for preferential CNS compartment involvement warrant further investigation, as they may inform CNS prophylaxis and treatment strategies. Leptomeningeal involvement is associated with poor prognosis, whereas ASCT shows a trend toward improved survival in treatment‐sensitive R‐SCNSL. Novel therapeutic approaches are needed to improve outcomes in R‐SCNSL, particularly for R‐SCNSL with leptomeningeal involvement.
Author Contributions
The methodology and conceptualization were developed collectively by G.S.N., H.W.T., and I.O. H.W.T was further responsible for supervision, visualization, project administration, and funding acquisition. I.O. was further responsible for the primary investigation, validation, data curation, formal analysis, and project administration. Z.L and L.J.W performed further formal analysis, validation, and software implementation. Resources were provided by I.O., Y.W., and T.M.H Additional investigation support was provided by R.L.K (Pathology Review), P.B.J., A.C.R., J.P., and T.E.W The writing ‐ original draft was prepared by I.O and H.W.T All authors contributed to writing – review and editing and approved the final manuscript for publication.
Conflicts of Interest
H.W.T. has served as a consultant for Curis Inc. G.S.N. reports receiving research funding from MorphoSys and Celgene, and has served as a consultant for Selvita, Karyopharm Therapeutics, Bristol Myers Squibb, Kite/Gilead, Genmab, Kymera, TG Therapeutics, MorphoSys, AbbVie, Debiopharm Group, Celgene, and Genentech. J.P. reports receiving research funding from Karyopharm, Biofourmis, and AstraZeneca, and has received honoraria from AbbVie and AstraZeneca. Y.W. reports receiving research funding (to institution) from Incyte, InnoCare, LOXO Oncology, Eli Lilly, MorphoSys, Novartis, Genentech, Genmab, AbbVie, BeiGene, and Merck; serving on advisory boards (compensation to institution) for Eli Lilly, LOXO Oncology, TG Therapeutics, Incyte, InnoCare, Kite, Jansen, BeiGene, AstraZeneca, Genmab, and AbbVie; serving as a consultant (compensation to institution) for InnoCare and AbbVie; and receiving honoraria (to institution) from Kite. T.M.H. reports receiving research funding from BMS/Celgene, GenMab, Sorrento, and Genentech, and has served as a consultant for Seagen and Eli Lilly. I.O., Z.L., L.J.W., P.B.J., A.C.R., R.L.K., and T.E.W. declare no competing interests.
Supporting information
Supporting Information S1
Acknowledgments
We gratefully acknowledge Dr. Thomas M. Habermann for his initiative and efforts in establishing the Mayo Clinic Lymphoma Database, which has been a tremendous resource for lymphoma research.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
- 1. Bernstein S. H., Unger J. M., Leblanc M., Friedberg J., Miller T. P., and Fisher R. I., “Natural History of CNS Relapse in Patients With Aggressive Non‐hodgkin's Lymphoma: A 20‐Year Follow‐Up Analysis of SWOG 8516 – The Southwest Oncology Group,” Journal of Clinical Oncology 27, no. 1 (2009): 114–119, 10.1200/jco.2008.16.8021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Bromberg J. E., Doorduijn J. K., Illerhaus G., et al. “Central Nervous System Recurrence of Systemic Lymphoma in the Era of Stem Cell transplantation–An International Primary Central Nervous System Lymphoma Study Group Project,” Haematologica 98, no. 5 (2013): 808–813, 10.3324/haematol.2012.070839. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. El‐Galaly T. C., Cheah C. Y., Bendtsen M. D., et al. “Treatment Strategies, Outcomes and Prognostic Factors in 291 Patients With Secondary CNS Involvement by Diffuse Large B‐Cell Lymphoma,” European Journal of Cancer 93 (2018): 57–68, 10.1016/j.ejca.2018.01.073. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Ferreri A. J., Donadoni G., Cabras M. G., et al. “High Doses of Antimetabolites Followed by High‐Dose Sequential Chemoimmunotherapy and Autologous Stem‐Cell Transplantation in Patients With Systemic B‐Cell Lymphoma and Secondary CNS Involvement: Final Results of a Multicenter Phase II Trial,” Journal of Clinical Oncology 33, no. 33 (2015): 3903–3910, 10.1200/jco.2015.61.1236. [DOI] [PubMed] [Google Scholar]
- 5. Ferreri A. J. M., Doorduijn J. K., Re A., et al. “MATRix‐RICE Therapy and Autologous Haematopoietic Stem‐cell Transplantation in Diffuse Large B‐Cell Lymphoma With Secondary CNS Involvement (MARIETTA): An International, Single‐Arm, Phase 2 Trial,” Lancet Haematology 8, no. 2 (2021): e110–e121, 10.1016/s2352-3026(20)30366-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Klanova M., Sehn L. H., Bence‐Bruckler I., et al. “Integration of Cell of Origin Into the Clinical CNS International Prognostic Index Improves CNS Relapse Prediction in DLBCL,” Blood 133, no. 9 (2019): 919–926, 10.1182/blood-2018-07-862862. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Alderuccio J. P., Khwaja J., Han S., et al. “Prognostication and Treatment‐Related Outcomes in Secondary Central Nervous System Involvement of Large B‐Cell Lymphoma (SCNSL): Evidence From a Large International Cohort,” Blood 144, no. 1 (2024): 571, 10.1182/blood-2024-199781. [DOI] [Google Scholar]
- 8. Baggio D., Han S., Ghione P., et al. “339 | Outcomes in an International Multicentre Study Of 600 Patients With Cns Relapse of Large B Cell Lymphoma: Compartment Of Relapse Is Prognostic,” Hematological Oncology 43, no. 3 (2025): e339_70094, 10.1002/hon.70094_339. [DOI] [Google Scholar]
- 9. Boehme V., Zeynalova S., Kloess M., et al. “Incidence and Risk Factors of Central Nervous System Recurrence in Aggressive lymphoma–A Survey of 1693 Patients Treated in Protocols of the German High‐Grade Non‐Hodgkin's Lymphoma Study Group (DSHNHL),” Annals of Oncology 18, no. 1 (2007): 149–157, 10.1093/annonc/mdl327. [DOI] [PubMed] [Google Scholar]
- 10. Savage K. J., Slack G. W., Mottok A., et al. “Impact of Dual Expression of MYC and BCL2 by Immunohistochemistry on the Risk of CNS Relapse in DLBCL,” Blood 127, no. 18 (2016): 2182–2188, 10.1182/blood-2015-10-676700. [DOI] [PubMed] [Google Scholar]
- 11. Schmitz N., Zeynalova S., Nickelsen M., et al. “CNS International Prognostic Index: A Risk Model for CNS Relapse in Patients With Diffuse Large B‐Cell Lymphoma Treated With R‐CHOP,” Journal of Clinical Oncology 34, no. 26 (2016): 3150–3156, 10.1200/jco.2015.65.6520. [DOI] [PubMed] [Google Scholar]
- 12. Sehn L. H. and Salles G., “Diffuse Large B‐Cell Lymphoma,” New England Journal of Medicine 384, no. 9 (2021): 842–858, 10.1056/nejmra2027612. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Haldorsen I. S., Espeland A., and Larsson E. M., “Central Nervous System Lymphoma: Characteristic Findings on Traditional and Advanced Imaging,” AJNR Am J Neuroradiol 32, no. 6 (2011): 984–992, 10.3174/ajnr.a2171. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Malikova H., Burghardtova M., Koubska E., Mandys V., Kozak T., and Weichet J., “Secondary Central Nervous System Lymphoma: Spectrum of Morphological MRI Appearances,” Neuropsychiatr Dis Treat 14 (2018): 733–740, 10.2147/ndt.s157959. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Wu Y., Wang Y., Sun X., et al. “Parenchymal Central Nervous System Involvement in Aggressive B‐Cell Lymphoma: Retrospective Analysis of Clinical and MRI Features in a Chinese Population,” BMC Neurology 19, no. 1 (2019): 268, 10.1186/s12883-019-1511-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Grommes C. and DeAngelis L. M., “Primary CNS Lymphoma,” Journal of Clinical Oncology 35, no. 21 (2017): 2410–2418, 10.1200/jco.2017.72.7602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Al F. K. A., Fang P., Gule‐Monroe M., et al. “Prognostic Value of Disease Distribution in Secondary Central Nervous System Diffuse Large B Cell Lymphoma Treated With Radiation Therapy,” Leukemia and Lymphoma 62, no. 10 (2021): 2400–2407, 10.1080/10428194.2021.1919656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Meng X., Yu J., Fan Q., et al. “Characteristics and Outcomes of Non‐Hodgkin's Lymphoma Patients With Leptomeningeal Metastases,” International Journal of Clinical Oncology 23, no. 4 (2018): 783–789, 10.1007/s10147-018-1268-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Suzuki T., Fujii K., Matsunaga N., et al. “Impact of MYD88 and/or CD79B Mutations on Central Nervous System Relapse in Patients With Diffuse Large B‐Cell Lymphoma,” British Journal of Haematology 207, no. 1 (2025): 92–100, 10.1111/bjh.20099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Bleyer W. A. and Poplack D. G., “Intraventricular Versus Intralumbar Methotrexate for Central‐Nervous‐System Leukemia: Prolonged Remission With the Ommaya Reservoir,” Medical and Pediatric Oncology 6, no. 3 (1979): 207–213, 10.1002/mpo.2950060304. [DOI] [PubMed] [Google Scholar]
- 21. Chamberlain M. C., Kormanik P. A., and Barba D., “Complications Associated With Intraventricular Chemotherapy in Patients With Leptomeningeal Metastases,” Journal of Neurosurgery 87, no. 5 (1997): 694–699, 10.3171/jns.1997.87.5.0694. [DOI] [PubMed] [Google Scholar]
- 22. Mason W. P., Yeh S. D., and DeAngelis L. M., “111Indium‐Diethylenetriamine Pentaacetic Acid Cerebrospinal Fluid Flow Studies Predict Distribution of Intrathecally Administered Chemotherapy and Outcome in Patients With Leptomeningeal Metastases,” Neurology 50, no. 2 (1998): 438–444, 10.1212/wnl.50.2.438. [DOI] [PubMed] [Google Scholar]
- 23. Kroin J. S., “Intrathecal Drug Administration. Present Use and Future Trends,” Clinical Pharmacokinetics 22, no. 5 (1992): 319–326, 10.2165/00003088-199222050-00001. [DOI] [PubMed] [Google Scholar]
- 24. Shapiro W. R., Young D. F., and Mehta B. M., “Methotrexate: Distribution in Cerebrospinal Fluid After Intravenous, Ventricular and Lumbar Injections,” New England Journal of Medicine 293, no. 4 (1975): 161–166, 10.1056/nejm197507242930402. [DOI] [PubMed] [Google Scholar]
- 25. Atkinson A. J. Jr, “Intracerebroventricular Drug Administration,” Transl Clin Pharmacol 25, no. 3 (2017): 117–124, 10.12793/tcp.2017.25.3.117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Katayama M., Ohgino K., and Tamura R., Leptomeningeal Metastases and Intrathecal Chemotherapy, Digital Medicine and Health Technology, Vol. 2022 (2022), 1–23, 10.5772/dmht.03. [DOI] [Google Scholar]
- 27. National Comprehensive Cancer N . Central Nervous System Cancers, (2025).
- 28. Arshad N., Biswas N., Gill J., Kesari S., and Ashili S., “Drug Delivery in Leptomeningeal Disease: Navigating Barriers and Beyond,” Drug Delivery 31, no. 1 (2024): 2375521, 10.1080/10717544.2024.2375521. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Nishimura N., Takeuchi K., Asaka R., et al. “MYD88 L265P Mutation Detected by Digital PCR as a Prognostic Factor in Patients With Diffuse Large B‐Cell Lymphoma in Rituximab Era,” Leukemia Research 97 (2020): 106426, 10.1016/j.leukres.2020.106426. [DOI] [PubMed] [Google Scholar]
- 30. El‐Galaly T. C., Villa D., Michaelsen T. Y., et al. “The Number of Extranodal Sites Assessed by PET/CT Scan is a Powerful Predictor of CNS Relapse for Patients With Diffuse Large B‐Cell Lymphoma: An International Multicenter Study of 1532 Patients Treated With Chemoimmunotherapy,” European Journal of Cancer 75 (2017): 195–203, 10.1016/j.ejca.2016.12.029. [DOI] [PubMed] [Google Scholar]
- 31. Cui J.‐Y., Liu Y.‐F., Gao B., et al. “Model for Predicting Central Nervous System Relapse in Diffuse Large B‐Cell Lymphoma and Discussion of Prophylaxis Measures,” World Neurosurgery 179 (2023): e387–e396, 10.1016/j.wneu.2023.08.099. [DOI] [PubMed] [Google Scholar]
- 32. Gleeson M., Counsell N., Cunningham D., et al. “Central Nervous System Relapse of Diffuse Large B‐Cell Lymphoma in the Rituximab Era: Results of the UK NCRI R‐CHOP‐14 Versus 21 Trial,” Annals of Oncology 28, no. 10 (2017): 2511–2516, 10.1093/annonc/mdx353. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Qualls D. and Abramson J. S., “Advances in Risk Assessment and Prophylaxis for Central Nervous System Relapse in Diffuse Large B‐Cell Lymphoma,” Haematologica 104, no. 1 (2019): 25–34, 10.3324/haematol.2018.195834. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Saha A., Whiting J., Ammad‐ud‐din M., et al., “A Retrospective Study on Outcomes of Secondary CNS Lymphoma: Pattern of Relapse, Prognostic Factors, and Role of Consolidation Therapy,” Blood (2023). [Google Scholar]
- 35. Basso K. and Dalla‐Favera R., “BCL6: Master Regulator of the Germinal Center Reaction and Key Oncogene in B Cell Lymphomagenesis,” Advances in Immunology 105 (2010): 193–210, 10.1016/S0065-2776(10)05007-8. [DOI] [PubMed] [Google Scholar]
- 36. Lossos I. S., Jones C. D., Warnke R., et al. “Expression of a Single Gene, BCL‐6, Strongly Predicts Survival in Patients With Diffuse Large B‐Cell Lymphoma,” Blood 98, no. 4 (2001): 945–951, 10.1182/blood.v98.4.945. [DOI] [PubMed] [Google Scholar]
- 37. Doorduijn J. K., van Imhoff G. W., van der Holt B., et al. “Treatment of Secondary Central Nervous System Lymphoma With Intrathecal Rituximab, High‐Dose Methotrexate, and R‐DHAP Followed by Autologous Stem Cell Transplantation: Results of the HOVON 80 Phase 2 Study,” Hematological Oncology 35, no. 4 (2017): 497–503, 10.1002/hon.2342. [DOI] [PubMed] [Google Scholar]
- 38. Agnieszka K., Thomas E., Eckhard T., et al., “Phase II Study of Central Nervous System (CNS)‐Directed Chemotherapy Including High‐Dose Chemotherapy With Autologous Stem Cell Transplantation for CNS Relapse of Aggressive Lymphomas,” Haematologica 98, no. 3 (2013): 364–370, 10.3324/haematol.2012.077917. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting Information S1
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.