Rituximab With High-Dose Methotrexate In Newly Diagnosed Primary Central Nervous System Lymphoma: a Single-Center Experience From China

Induction chemotherapy based on high-dose methotrexate is considered as the standard approach for newly diagnosed primary central nervous system lymphomas (PCNSLs). However, the best combination chemotherapeutic regimen remains unclear. This study aimed to determine the ecacy and toxicities of rituximab with methotrexate (R-M regimen). Consecutive 37 Chinese patients receiving R-M regimen as induction chemotherapy were retrospectively identied from January 2015 to June 2020 from our center in eastern China. Fourteen patients receiving rituximab plus methotrexate with cytarabine (R-MA regimen) at the same period were identied as the positive control group. The response rates, survival, toxicities, length of hospital stay (LOS), and cost were compared. Compared with the R-MA regimen, the R-M regimen showed comparable response rate and survival outcomes, but had fewer grade 3-4 hematological toxicities, shorter LOS, lower mean total hospitalization cost and lower mean total antibiotic cost. Overall response after two cycles of chemotherapy, complete remission at the end of induction chemotherapy and ECOG>3 were independent prognostic factors for overall survival. In conclusion, R-M regimen is an effective and well-tolerated combination treatment for PCNSLs, which warrants further evaluation in randomized trials.

Introduction Primary central nervous system lymphoma (PCNSLs) is a rare and aggressive extranodal non-Hodgkin lymphoma located in the brain, leptomeninges, spinal cord, eyes or cerebrospinal uid (CSF) 1 . PCNSLs are characterized by short-lasting responses and poor prognosis. High-dose methotrexate (HD-MTX) is the standard treatment for newly diagnosed PCNSLs. However, the e cacy of HD-MTX monotherapy is unsatisfactory with a median overall survival (OS) of 16-38 months and progression-free survival (PFS) of 4-10 months [2][3][4] . As such, other drugs in combination with HD-MTX have been investigated to improve the outcome 5 . Cytarabine (Ara-C) is a cell cycle-phase speci c antimetabolite, commonly used in chemotherapy for hematologic malignancies. IELSG20, a phase 2 trial, indicated that the combination of Ara-C with HD-MTX increased the CR rate (46% vs. 18%) and improved the 3-year OS (46% vs. 32%) in PCNSLs 4 . IELSG32, another phase 2 trial, revealed that patients receiving the HD-MTX plus Ara-C (MA regimen) together with thiotepa and rituximab (MATRix) had a higher CR rate (49% vs. 23%) and improved PFS and OS, than those receiving the MA regimen alone 6 . Although combination chemotherapeutic regimens bring increasing e cacy, hematological toxicities are the primary concerns. Therefore, the best combination chemotherapeutic regimen remains unclear.
Rituximab, a highly speci c monoclonal antibody against CD20, is widely used to treat diffuse large B cell lymphoma (DLBCL). The combination of rituximab with the cyclophosphamide, doxorubicin, vincristine and prednisone (R-CHOP) regimen has been proved to signi cantly improve the OS of systemic DLBCL 7 . Although rituximab's clinical effect in PCNSLs remains controversial due to its large size and poor penetration of the blood-brain barrier (BBB), several preliminary studies suggest that adding rituximab to cytotoxic chemotherapy is safe and effective for PCNSLs 2,8,9 . Studies have indicated that rituximab is likely to reach therapeutic concentrations and can induce responses in contrast-enhanced lesions, in which there is a substantial disruption of the BBB 10,11 . Two randomized controlled trials were designed to formally determine the effects of rituximab on PCNSLs. The IELSG32 study stated that the addition of rituximab to the MA regimen (R-MA regimen) was more likely to induce a response than the MA regimen alone (ORR,74% vs. 53%) and there was a trend for improved PFS (pvalue 0.051) and OS (p-value 0.095) with overall minimal added toxicities 6 . However, the HOVON 105/ALLG NHL 24 study, a phase 3 randomized study, did not nd signi cantly increased PFS, or OS from the addition of rituximab to the MTX, carmustine, teniposide, and prednisolone (MBVP) chemotherapy regimen compared with the MBVP regimen alone in newlydiagnosed PCNSLs 12 , which was contrary to the results in the IELSG32 study. Interestingly, subgroup analysis in the HOVON105 study showed that the application of rituximab might bene t patients under 60 years of age. Even with the con icting results of the above two prospective randomized studies, given the relatively low toxicities and single-agent activity of rituximab in relapsed/refractory PCNSLs, rituximab has been widely used in clinical practice and incorporated into clinical trials 13 .
To balance therapy intensi cation with side effects in our center, rituximab with methotrexate (R-M regimen) is widely applied for PCNSLs. This study aimed to retrospectively evaluate the e cacy and toxicities of the R-M regimen as induction chemotherapy for newly diagnosed PCNSLs.

Patients
From January 2015 to June 2020, consecutive patients with newly diagnosed PCNSLs, receiving the R-M regimen as induction therapy at the Second A liated Hospital, Zhejiang University, were identi ed from medical records. Given the inaccessibility of thiotepa in China, the R-MA regimen but not the MATRix regimen was widely applied since the IELSG32 results was rst published in 2016. Thus, patients receiving the R-MA regimen at the same period were identi ed as the positive control group. Patients ful lling the following criteria were enrolled in the study: newly diagnosed patients with PCNSL, de nite pathological diagnosis of DLBCL, and measurably enhanced lesions for response assessment. Patients with isolated ocular lymphomas or HIV-associated lymphoma were excluded. Informed consent was obtained from all patients enrolled in this study. According to the updated WHO classi cations, the diagnosis of PCNSLs was con rmed by two pathologists and consensus was reached for each patient. This study was approved by the Ethics Committee of the Second A liated Hospital, Zhejiang University.

Treatment protocol
Patients were scheduled to receive 6 cycles of induction chemotherapy on a 21-day cycle. Thereafter, consolidation chemotherapy with whole-brain radiation therapy (WBRT,30-36 Gy followed by a limited eld to gross disease to 45 Gy) was administered, depending on their age, nancial situation and willingness. Patients who failed to reach PR after 4 courses of induction chemotherapy or progressed at any time were withdrawn and received salvage treatment. The R-M regimen consisted of rituximab and MTX, while the R-MA regimen consisted of rituximab, MTX and Ara-C. Rituximab (375 mg/m2) was administered on day 0. HD-MTX (3.5 g/m2) was administered intravenously over 4 hours on day 1. Patients received adequate pre-and post-MTX hydration, urinary alkalization and leucovorin rescue. Leucovorin rescue was started 14 hours after completing HD-MTX infusion and administered at 15mg every 6 hours until serum MTX levels were ≤0.05 µmol/l. Ara-C (2 g/m2) was administered intravenously every 12 hours on days 2-3. All patients who had grade 3-4 neutropenia with infection received prophylactic lenograstim after each cycle and the dose of Ara-C was reduced thereafter.

Response evaluation
Treatment responses were evaluated every 2 induction chemotherapy cycles or at any time the neurological symptoms worsened or re-emerged by brain magnetic resonance imaging (MRI) with contrast enhancement. The responses were classi ed as complete remission (CR), uncon rmed complete remission (CRu), partial remission (PR), stable disease (SD), or progression disease (PD), according to the International Primary CNS Lymphoma Collaborative Group (IPCG) criteria 14 . For statistical analysis, CR and CRu de ned by the IPCG criteria were combined into a single "CR" category previously de ned by Response Evaluation Criteria in Solid Tumors in this study. Overall response=CR+PR. After completing the treatment, patients were assessed by repeated enhanced brain MRI every 3 months for the rst 2 years and every 6 months for the next 3 years.
OS was determined as the period from the date of diagnosis of PCNSLs to death from any cause or the last follow-up. PFS was de ned as the time from diagnosis to disease progression or death due to PCNSL or last follow-up.
Determination of length of hospital stay (LOS) LOS was calculated from the day of chemotherapy to discharge and recorded in whole days. When a case was readmitted for side effects, the duration was added to the initial LOS. LOS was calculated for the R-M group and R-MA group, respectively.

Cost analysis
The cost of hospital stays and antibiotics was determined from patient billing records and converted to US dollars at the currency exchange rate at the time of this study. The cost of hospital stays and antibiotics were calculated for the R-M group and R-MA group, respectively.

Statistical analysis
Continuous variables were summarized as means with 95% con dence intervals (95%CIs) and compared with an unpaired ttest. Categorical variables were summarized as frequencies with associated percentages. The patient characteristics and treatment responses of the two therapeutic groups were compared using the chi-square test or Fisher's exact test. The OS and PFS were calculated with the Kaplan-Meier method and compared with the log-rank test. The multivariate analysis for OS and PFS was performed based on the Cox proportional hazards regression model. The results were expressed as hazard ratios (HR) and their 95%CIs. Those factors with p-values less than 0.1 in univariate analysis were used to construct multivariate models for OS and PFS. For continuous variables, the best-identi ed cutoff point was determined by receiver operating characteristic (ROC) curve analysis. All statistical analyses were performed using SPSS v17.0 for Windows (SPSS Inc, Chicago, IL). An alpha value of two-sided p<0.05 was considered statistically signi cant.
All above methods were performed in accordance with the relevant guidelines and regulations.

Patient characteristics and treatment
A total of 34 patients receiving R-M regimen were enrolled. The diagnosis was achieved by surgery (40.5%), or stereotactic biopsy (59.5%), and no patients were diagnosed based on CSF. All PCNSLs were con rmed to be DLBCL. The median age of patients was 62 years (range, 27 to 80 years; 23 patients were>60 years old, and 14 were ≤60 years old). Five patients (13.5%) had elevated LDH above the upper limit of normal. 77.1% (27/35) patients had non-germinal center B cell-like (non-GCB) DLBCL. Severn (20.6%) patients were consolidated with WBRT.
Fourteen patients receiving the R-MA regimen as induction chemotherapy were identi ed as the positive control group. The characteristics of the 51 patients are described in Table 1. These patients were treated according to their age, nancial situation and willingness. The demographics and baseline clinical characteristics (sex, ECOG, LDH, deep structure involvement, biopsy type, Karnofsky performance status (KPS), CD10, C-Myc, Bcl-2, Bcl-6, MUM-1, pathological phenotype, and CSF protein) were not different between patients treated with the R-M and R-MA regimens, except for the age and number of lesions. Patients in the R-M group were older (p=0.006) than the R-MA group, and more patients in the R-M group had multiple lesions (p=0.024).

Toxicity and cost
Thirty-seven patients received a total of 184 cycles of the R-M regimen with a median of 6 cycles per patient (range, 1-6).
Twenty-four of the 37 patients completed the entire protocol. Only 1 patient (2.7%) was withdrawn due to severe myelosuppression with infection and 9 patients (24.3%) were withdrawn due to disease progression. The remaining 3 patients were withdrawn due to personal reasons. Fourteen patients received a total of 39 cycles of the R-MA regimen, with a median of 3 cycles per patient (range, 1-6). Only three of the 14 patients completed the entire protocol. Indeed, ve patients (35.7%) were withdrawn due to severe myelosuppression with infection, and six patients (42.9%) were withdrawn due to the disease progression. The total chemotherapy completion rate was higher in the R-M group than in the R-MA group (64.9% vs. 21.4%, p=0.006) ( Figure 1D). The chemotherapy interruption rate due to toxicities was signi cantly lower in the R-M group than in the R-MA group (2.7% vs. 35.7%, p=0.004).
The most frequent toxicities were hematologic toxicities and infection, as summarized in

Univariate and multivariate analyses of prognostic indicators
The results of univariate and multivariate analyses of prognostic indicators are summarized in Table 3. The univariate analysis indicated that response of overall response after two cycles of chemotherapy (p=0.000), complete remission at the end of the induction therapy (p=0.002) and overall response at the end of the induction therapy (p=0.000) and GCB subtype (p=0.019) were associated with increased PFS (Figure 2A-D). While overall response after two cycles of chemotherapy (p=0.003), complete remission at the end of the induction therapy (p=0.01), overall response at the end of the induction therapy (p=0.004), age>58 (p=0.041) and ECOG>3 (p=0.007) were associated with increased OS (Figure 2E-H). Disappointingly, we did not observe any association of LDH or deep brain involvement with PFS or OS. We incorporated all those factors with p-values<0.1 in univariate analysis into the multivariate Cox analysis and revealed that overall response after two cycles of induction chemotherapy (p=0.073),complete remission at the end of induction chemotherapy (p=0.070) and ECOG >3 (p=0.007) were independent prognostic factors for OS. While overall response at the end of induction chemotherapy (p=0.049) and multifocal (p=0.067) were identi ed as dependent prognostic indicators for PFS.

Discussion
We retrospectively evaluated the e cacy, toxicity and cost of the R-M regimen, and compared it with the R-MA regimen. We found that the response rates and survival were not signi cantly different between the two regimens, indicating that the e cacy of the R-M regimen was comparable to the R-MA regimen. However, the R-M regimen yielded a higher chemotherapy completion rate with fewer severe toxicities, shorter LOS and lower cost than the R-MA regimen for patients with newly diagnosed PCNSLs.
Induction chemotherapy based on HD-MTX is considered as the standard approach for newly diagnosed PCNSLs. However, the best chemotherapeutic regimen remains unclear. Combinations with other chemotherapeutic agents have been shown to improve response and survival in patients with PCNSL, although few randomized clinical trials have supported this practice 5 .
Controlled studies comparing the e cacy of different induction chemotherapy regimens were reviewed and are summarized in Appendix Table S1. However, comparisons were still challenging due to differing treatment schedules, consolidation strategies, and dosages and patient heterogeneity between the studies. It seems that adding more drugs into the HD-MTXbased chemotherapy regimens tends to improve the outcome but brings more toxicities. R-M regimen appears to be the combination chemotherapy with the least additional toxicities. Holdhoff et al. 2 revealed that compared with the single HD-MTX group, the R-M regimen improved CR rates and survival. In contrast, Kansara et al. 3 reported that the addition of Rituximab to HD-MTX did not appear to improve outcomes. Both studies indicate that the addition of rituximab was feasible and well tolerated with overall minimal added toxicities. Our data suggest that the e cacy of the R-M regimen is comparable to that of the R-MA regimen, and most importantly, the R-M regimen shows the advantage of a better toxicity pro le, although the sample size of the current study was small.
The combination of HD-Ara-C with HD-MTX has signi cantly improved outcomes, as described in large studies 4,22 . After the IELSG20 results were rst published in 2009 4 , HD-Ara-C was widely applied in the management of PCNSLs. Recently, several studies reported the disappointing results of the HD-Ara-C containing regimens. One retrospective study demonstrated that neither PFS nor OS was higher in the Ara-C group than a therapeutic regimen without Ara-C 18 . A randomized clinical trial (RCT) by Wu et al. 15 showed that the non-Ara-C group (receiving the fotemustine, teniposide, and dexamethasone (FTD) regimen) had better apparent effects and safety than the Ara-C group (receiving the MA regimen). Hematological toxicities were the main concern. As expected, grade 3-4 hematological toxicities including neutropenia (90%, 56% and 59%, respectively), thrombocytopenia (92%, 74%and 59%, respectively) and anemia (46%, 36% and 23%, respectively) were quite common in the IELSG20 trial, IELSG32 trial and the current study. Chemotherapy interruption due to toxicities occurred in 17.9%, 12% and 35.7% of patients in the IELSG20, IELSG32 and the current study, respectively. What's worse, a treatmentrelated mortality of 6-9% was observed in these clinical trials. Thus, more effective and less toxic therapeutic regimens are urgently needed to balance therapy intensi cation with side effects, especially for old and weak patients. Our data suggest that the e cacy of the R-M regimen is comparable to that of the R-MA regimen with much fewer toxicities. We considered that the high frequency of grade 3-4 side effects with Ara-C might compromise the OS bene t. These toxicities were the main contributors to cost and prolonged hospital stay, resulting in a massive nancial burden. In recent years, newly developed targeted drugs may provide new treatment options in PCNSL patients 23 . Ibrutinib (a selective covalent BTK inhibitor) 24 , nivolumab 25 (a PD1 block) and lenalidomide (an immunomodulatory imide drug) 26 , showed promising response rates in refractory/recurrent PCNSL patients with good tolerability. we initiated a phase-2 trial, comparing the e cacy and safety of the R-M regimen with or without lenalidomide for newly-diagnosed PCNSLs, is ongoing (NCT04481815). As a result, we believe that in the rituximab era, HD-Ara-C may be eliminated from the initial combination chemotherapy regimen for PCNSLs, which warrants further evaluation in randomized trials.
There are two prognostic scoring systems developed speci cally for PCNSLs. One is the International Extranodal Lymphoma Study Group (IELSG) scoring system, which divides patients into three different risk classes based on age, ECOG performance status, LDH, CSF total protein concentration, and involvement of deep brain structures 27 . The other is the Memorial Sloan-Kettering Cancer Center (MSKCC) scoring system, which divides patients into three different risk classes, using age and KPS 28 . In our retrospective analysis, only ECOG>3 was predictive of prognosis in multivariate analyses. Interestingly, we found that the early response and nal response could be used for patient prognostication. Patients who achieved overall response within two cycles of therapy had an improved OS compared with those who did not, indicating a need for early treatment modi cation for those without early response in clinical practice. Patients who did not achieve CR at the end of therapy had a poor OS compared with those who did, indicating that treatment should be resumed for those without CR with consolidation (such as WBRT or autologous bone marrow transplantation) at the end of treatment.
Our study has several limitations given that it is a retrospective, single-institution analysis with modest sample size. First, the CSF protein data of 33% of patients were not available for the current study, similar to other retrospective studies. Therefore, it was impossible to determine different responses among IELSG subgroups. Second, it should be highlighted that only 3 of the 14 enrolled patients (21.4%) in the R-MA group completed the entire protocol, which limited true judgments of its effectiveness. However, our study has multiple strengths. First, all our patients had a uniform, identical pathological diagnosis and received standardized treatment schedules, despite this being a retrospective study. Second, our singleinstitution retrospective study has one of the largest samples of patients with PCNSLs.

Conclusion
In conclusion, the e cacy of the R-M regimen is comparable to that of the R-MA regimen, but with fewer toxicities, well tolerability, shorter LOS and lower cost. Our results suggest that R-M regimen is an effective and well-tolerated combination treatment for PCNSLs, which warrants further evaluation in randomized trials.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableS1.SummaryoftheinductionchemotherapyregimensusedfornewlydiagnosedpatientswithPCNSLs.docx