A Systematic Review of High-Dose Methotrexate for Adults with Primary Central Nervous System Lymphoma

Simple Summary High-dose methotrexate (HDMTX) is the backbone of induction therapy for primary central nervous system lymphoma (PCNSL). There are numerous different protocols to treat PCNSL that incorporate a wide range of HDMTX doses and various combinations with other chemotherapeutic agents. This systematic review was conducted to summarize the various treatment regimens for PCNSL and determine outcomes among the different doses of HDMTX and combination regimens. The findings are intended to provide guidance on the optimal dose and regimen of HDMTX for the treatment of PCNSL. Abstract Primary central nervous system lymphoma (PCNSL) is a highly aggressive non-Hodgkin lymphoma that is confined within the CNS. Due to its ability to cross the blood–brain barrier, high-dose methotrexate (HDMTX) is the backbone for induction chemotherapy. This systematic review was conducted to observe outcomes among different HDMTX doses (low, <3 g/m2; intermediate, 3–4.9 g/m2; high, ≥5 g/m2) and regimens used in the treatment of PCNSL. A PubMed search resulted in 26 articles reporting clinical trials using HDMTX for PCNSL, from which 35 treatment cohorts were identified for analysis. The median dose of HDMTX used for induction was 3.5 g/m2 (interquartile range IQR, 3–3.5); the intermediate dose was most frequently used in the studies examined (24 cohorts, 69%). Five cohorts used HDMTX monotherapy, 19 cohorts used HDMTX + polychemotherapy, and 11 cohorts used HDMTX + rituximab ± polychemotherapy. Pooled overall response rate (ORR) estimates for low, intermediate, and high dose HDMTX cohorts were 71%, 76%, and 76%, respectively. Pooled 2-year progression-free survival (PFS) estimates for low, intermediate, and high HDMTX dose cohorts were 50%, 51%, and 55%, respectively. Regimens that included rituximab showed a tendency to have higher ORR and 2-year PFS than those that did not include rituximab. These findings indicate that current protocols utilizing 3–4 g/m2 of HDMTX in combination with rituximab provide therapeutic efficacy in PCNSL.


Introduction
Primary central nervous system lymphoma (PCNSL) is a highly aggressive non-Hodgkin lymphoma that is confined to the central nervous system (CNS) and vitreoretinal space. It is a rare malignancy that accounts for 4% of intracranial neoplasms and 4-6% of extra-nodal lymphomas, and can occur in both immunocompromised and immunocompetent individuals [1,2]. More than 90% of PCNSLs are of diffuse large B-cell lymphoma (DLBCL) phenotype, and 'primary DLBCL of the CNS' was recognized as a distinct entity by the 2017 World Health Organization (WHO) classification of hematopoietic and lymphoid tumors [3,4]. PCNSL predominantly affects adults older than 60 years of age (median

Methods
The selection and systematic review of appropriate studies was performed in accordance to the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) 2020 statement guidelines ( Figure 1) [19]. This systematic review was part of a broader review investigating the optimal use of HDMTX in CNS tumors. A Medline/PubMed search was conducted for papers that were published up to July 2021 using Medical Subject Heading (MeSH) search terms determined from the 2021 World Health Organization (WHO) Classification of Tumors of the Central Nervous System [20]. Criteria for inclusion were clinical studies on CNS tumors in humans and use of HDMTX (defined as dose ≥ 500 mg/m 2 ). Criteria for exclusion were nonclinical or animal studies and review articles. The initial search identified 587 articles pertaining to the use of HDMTX in CNS malignancies, all of which were written in English, Spanish, or German. Following removal of duplications, four authors (MG, GV, CS, PS) independently assessed article eligibility based on review of abstracts and came to a consensus on the selection of 264 articles. To focus solely on PCNSL, inclusion criteria were refined to include only prospective clinical trials for PCNSL, and retrospective studies, case series, long-term follow-up studies, and articles reporting results on less than 25 patients were excluded. Ultimately, 26 full-text articles of prospective clinical trials reporting the use of HDMTX in PCNSL were reviewed for this analysis. studies, and articles reporting results on less than 25 patients were excluded. Ultimately, 26 full-text articles of prospective clinical trials reporting the use of HDMTX in PCNSL were reviewed for this analysis.

Data Collection
Each arm of every randomized trial was identified as a separate cohort of uniformly treated patients, resulting in 35 analytic cohorts extracted from the 26 prospective clinical trials. Information was collected from each cohort to include study type, number of patients, tumor histology (e.g., DLBCL), level of evidence, methotrexate dose, number of cycles and courses, use of rituximab, use of other chemotherapeutic agents for induction, chemotherapy to the CNS compartment (i.e., via intrathecal or intracerebroventricular administration) during induction, and type of consolidation therapy. Additional data were collected on treatment-related toxicities. Outcomes of interest were 2-year progression free survival (PFS) and ORR, where ORR included patients with complete response (CR) or partial response (PR) at the end of induction. In cohorts for which PFS was not specifi-

Data Collection
Each arm of every randomized trial was identified as a separate cohort of uniformly treated patients, resulting in 35 analytic cohorts extracted from the 26 prospective clinical trials. Information was collected from each cohort to include study type, number of patients, tumor histology (e.g., DLBCL), level of evidence, methotrexate dose, number of cycles and courses, use of rituximab, use of other chemotherapeutic agents for induction, chemotherapy to the CNS compartment (i.e., via intrathecal or intracerebroventricular administration) during induction, and type of consolidation therapy. Additional data were collected on treatment-related toxicities. Outcomes of interest were 2-year progression free survival (PFS) and ORR, where ORR included patients with complete response (CR) or partial response (PR) at the end of induction. In cohorts for which PFS was not specifically reported, PFS data were estimated from the survival curves that were presented or supplementary data.
Proportional meta-analyses were applied to estimate the pooled effect of different HDMTX protocols on ORR and 2-year PFS. As heterogeneity among the included studies was expected due to methodological differences, a random effects model was applied; heterogeneity was assessed using a chi-squared heterogeneity test and I-squared statistic. Forest plots were used to display pooled estimates and individual study results in selected protocols. To assess the impact of rituximab on the efficacy of induction therapy for PCNSL, exploratory comparisons of ORR and 2-year PFS estimates by HDMTX dosage categories between cohorts that received rituximab (Group 3) and those that did not receive rituximab (Groups 1 and 2) were conducted.

Results
From the 26 articles of prospective clinical trials 35 cohorts were identified; a total of 2115 patients comprising the 35 cohorts had received induction with HDMTX and were included in the analysis. Table 1 summarizes the clinical trials and characteristics of each analytical cohort as identified by unique cohort numbers.

Characteristics of the Analytic Cohorts
All 35 cohorts were comprised of adult patients with PCNSL; 30 cohorts included patients older than 65 years of age. Histologically, DLBCL was the most common phenotype accounting for 57% of the study population while other subtypes (e.g., Burkitt lymphoma, T-cell lymphoma) were reported in 2%; there was no histological subtype reported in 41%. Intra-ocular tumors were reported in 55 patients, and 139 patients had leptomeningeal dissemination at diagnosis.
ORR estimates for cohorts that received <5 courses and ≥5 courses of HDMTX were 73% and 79%, respectively. Cohorts that included CNS chemotherapy as part of induction therapy had a pooled estimate ORR of 79% compared to 73% in those that did not administer CNS chemotherapy (Table 2).   Pooled ORR estimates of cohorts that received rituximab and those that did not wer analyzed separately among the different HDMTX dosage categories. None of the low dos HDMTX cohorts received rituximab. For those receiving intermediate dose HDMTX (3 4.9 g/m 2 ), the ORR of 86% [95% CI, 81-90%] in cohorts that received rituximab was highe than in cohorts that did not receive rituximab, at 69% [95% CI, 56-79%]. Among the hig dose HDMTX cohorts, the ORR was 78% [95% CI, 38-100%] in those that received rituxi mab and 72% [95% CI, 61-82%] for those that did not receive rituximab (Figure 4). Pooled ORR estimates of cohorts that received rituximab and those that did not were analyzed separately among the different HDMTX dosage categories. None of the low dose HDMTX cohorts received rituximab. For those receiving intermediate dose HDMTX (3-4.9 g/m 2 ), the ORR of 86% [95% CI, 81-90%] in cohorts that received rituximab was higher than in cohorts that did not receive rituximab, at 69% [95% CI, 56-79%]. Among the high dose HDMTX cohorts, the ORR was 78% [95% CI, 38-100%] in those that received rituximab and 72% [95% CI, 61-82%] for those that did not receive rituximab (Figure 4).  The pooled 2-year PFS estimates for cohorts that received <5 courses and ≥5 courses of HDMTX were 50% and 52%, respectively. Cohorts that included CNS chemotherapy as part of induction therapy had a pooled 2-year PFS estimate of 52%, while it was 51% for those that did not receive CNS chemotherapy. (Table 2).  The pooled 2-year PFS estimates for cohorts that received <5 courses and ≥5 courses of HDMTX were 50% and 52%, respectively. Cohorts that included CNS chemotherapy as part of induction therapy had a pooled 2-year PFS estimate of 52%, while it was 51% for those that did not receive CNS chemotherapy. (Table 2).    Although there were four total cohorts that used high dose HDMTX, i.e., two in the ritux imab (+) group and two in the rituximab (−) groups, 2-year PFS data was missing from one cohort from the rituximab (−) group.   (Figure 7). Although there were four total cohorts that used high dose HDMTX, i.e., two in the rituximab (+) group and two in the rituximab (−) groups, 2-year PFS data was missing from one cohort from the rituximab (−) group.

Consolidation Therapy
Consolidation therapy was provided in 29 of the 35 cohorts. Radiation therapy (RT) only was given in 11 cohorts, high-dose chemotherapy (HDCT) only was given in 4 cohorts, stem cell transplantation (SCT) only was used in 2 cohorts, RT + HDCT in 6 cohorts, RT + SCT in 1 cohort, and SCT or RT in 3 cohorts. Cytarabine was used for HDCT consolidation in most cohorts; HDMTX was given in cohort number 125 [21], temozolamide was given in cohort number 130 [30], and cytarabine + etoposide was administered in cohort number 165 [42] (Table 1). RT dosage ranged from 25 to 54 Gy; pooled estimates for ORR and 2-year PFS appear similar in cohorts that did and did not receive RT (Table 2). In the 6 cohorts that received SCT (alone or in combination), ORR and 2-year PFS showed a tendency to be lower than the cohorts that did not receive SCT (Table 2).

Toxicities
Treatment-induced toxicities were reported in 32 of the 35 cohorts while the remaining 3 cohorts provided no information regarding toxicities. Renal toxicity was most common with 30 cohorts reporting its occurrence, among which 22 cohorts reported grade 3-4 toxicity and 5 reported grade 1-2; severity was not specified in 3 cohorts. Neurotoxicity was reported in 18 cohorts, among which 10 cohorts reported grade 3-4 neurotoxicity. Although the exact timing of its occurrence (i.e., after induction with HDMTX or after consolidation therapy) was not always specified, 9 cohorts reporting neurotoxicity included IT or ICV chemotherapy and 14 cohorts reported RT as a component of consolidation. Mucositis was reported in 15 cohorts, all of which were of grade 3-4 severity.

Consolidation Therapy
Consolidation therapy was provided in 29 of the 35 cohorts. Radiation therapy (RT) only was given in 11 cohorts, high-dose chemotherapy (HDCT) only was given in 4 cohorts, stem cell transplantation (SCT) only was used in 2 cohorts, RT + HDCT in 6 cohorts, RT + SCT in 1 cohort, and SCT or RT in 3 cohorts. Cytarabine was used for HDCT consolidation in most cohorts; HDMTX was given in cohort number 125 [21], temozolamide was given in cohort number 130 [30], and cytarabine + etoposide was administered in cohort number 165 [42] (Table 1). RT dosage ranged from 25 to 54 Gy; pooled estimates for ORR and 2-year PFS appear similar in cohorts that did and did not receive RT (Table 2). In the 6 cohorts that received SCT (alone or in combination), ORR and 2-year PFS showed a tendency to be lower than the cohorts that did not receive SCT (Table 2).

Toxicities
Treatment-induced toxicities were reported in 32 of the 35 cohorts while the remaining 3 cohorts provided no information regarding toxicities. Renal toxicity was most common with 30 cohorts reporting its occurrence, among which 22 cohorts reported grade 3-4 toxicity and 5 reported grade 1-2; severity was not specified in 3 cohorts. Neurotoxicity was reported in 18 cohorts, among which 10 cohorts reported grade 3-4 neurotoxicity. Although the exact timing of its occurrence (i.e., after induction with HDMTX or after consolidation therapy) was not always specified, 9 cohorts reporting neurotoxicity included IT or ICV chemotherapy and 14 cohorts reported RT as a component of consolidation.
Mucositis was reported in 15 cohorts, all of which were of grade 3-4 severity.

Discussion
This systematic review aimed to summarize the treatment protocols and outcomes of clinical trials for PCNSL with respect to dose of HDMTX and different regimens of HDMTX (i.e., HDMTX monotherapy, HDMTX + polychemotherapy, and HDMTX + rituximab ± polychemotherapy). Twenty-six articles reporting on clinical trials using HDMTX were reviewed, from which 35 treatment cohorts were identified and used for analysis.
The dose of HDMTX used in protocols ranged widely from 1 to 8 g/m 2 , with a median dose of 3 g/m 2 . Based on this median value and the doses used across the 35 cohorts (Table 1), HDMTX dose categories were defined as low dose (<3 g/m 2 ), intermediate dose (3 to 4.9 g/m 2 ), and high dose (≥5 g/m 2 ) to conduct analysis. (As HDMTX was administered intravenously in all cohorts, peak plasma methotrexate levels were proportional to the HDMTX dose [43]. This suggests that outcomes are closely related to the dose of methotrexate.) The intermediate dose was most frequently used as it was reported in 24 of the 35 cohorts (69%); however, pooled ORR estimates were 71%, 76%, and 76% for the low, intermediate, and high dose cohorts, respectively, showing no difference among HDMTX dose categories (Figure 2). Similarly, there was no difference in the pooled estimates for 2-year PFS among the dose categories: 50% (low dose), 51% (intermediate dose), and 55% (high dose) (Figure 3). These findings suggest that higher doses of HDMTX may not be necessary and the current mainstay of 3 to 3.5 g/m 2 is sufficient in achieving treatment efficacy while reducing the risk of severe toxicities that may arise from use of higher doses, especially in individuals with reduced renal function [44,45]. The publication dates for the articles included in this review span a period between 1992 (DeAngelis et al.) [23] and 2021 (Fu et al.) [28]. Over the nearly 30 years that encompass this period, there was no significant linear trend found between HDMTX dose and the chronological passage of time; most regimens used 3-3.5 g/m 2 reflecting the fact that this dose has been and continues to be efficacious and safe in the treatment of PCNSL. The relatively smaller number of cohorts that used <3 g/m 2 (n = 7) or ≥ 5g/m 2 (n = 4) of HDMTX in this review, however, limited the ability to fully assess the impact that lower or higher doses of HDMTX might have on treatment efficacy. In a study determining the optimal dose of HDMTX in 50 patients with PCNSL, Dalia et al. reported no significant difference in PFS or OS when comparing HDMTX doses of 8 g/m 2 to 3.5 g/m 2 , which was concordant with our results [46]. They also reported that neither HDMTX dose reductions or higher cumulative HDMTX doses were associated with significant differences in PFS or OS, which further support our results. Contrary to these findings, in a comparison of outcomes using 3.5 g/m 2 (n = 32) and 8 g/m 2 (n = 41) in patients with PCNSL, Li et al. reported significantly greater rates of CR (68.3% vs 43.8%, p = 0.03) and longer median PFS (17 months vs 9 months, p = 0.03) in patients receiving 8 g/m 2 compared to those receiving 3.5 g/m 2 [47]. These results, however, should be interpreted with caution as the median age and IQR were significantly lower in the 8 g/m 2 group at 49 years [IQR 42-55] compared to the 3.5 g/m 2 group at 61 years [IQR 51-69] (p = 0.01), as age is a significant prognostic factor for PCNSL [48]. Li et al. did further compare outcomes between the two dose categories only in patients younger than 65 years of age and found a higher median PFS of 17.7 months in the 8 g/m 2 group compared to 7 months in the 3.5 g/m 2 group (p = 0.02).
To further examine the impact of rituximab, exploratory comparisons of ORR and 2-year PFS were conducted between cohorts that received rituximab (regimen Group 3) and those that did not receive rituximab (regimen Groups 1 and 2) among the different HDMTX dose categories. Among cohorts using the intermediate dose (3-4.9 g/m 2 ), pooled ORR estimates were 86% in cohorts that received rituximab which was higher than the 69% found in cohorts that did not receive rituximab (Figure 4). Among cohorts using high dose HDMTX (≥ 5 g/m 2 ), pooled ORR estimates were 73% for those receiving rituximab and 72% for those that did not receive rituximab. The small number of cohorts that used high dose HDMTX in both groups (two cohorts each), however, likely did not provide adequate power to fully represent any potential effect the higher dose might have had on ORR. A similar pattern was found when comparing pooled 2-year PFS estimates among the cohorts that used the intermediate dose (3-4.9 g/m 2 ): 58% in cohorts that received rituximab and 45% in those that did not receive rituximab (Figure 7). Although statistical comparisons could not be performed due to the lack of patient level data, the higher ORR and 2-year PFS in cohorts that received rituximab imply that the addition of rituximab to HDMTX-based induction regimen may positively impact treatment outcomes of PCNSL. To date there have been two prospective randomized studies investigating the efficacy of rituximab in PCNSL: the IELSG32 trial [15] comparing three arms (A, HDMTX + cytarabine; B, HDMTX + cytarabine + rituximab; C, HDMTX + cytarabine + rituximab + thiotepa) and the HOVON105/ALLG NHL24 trial [14] which compared MBVP with and without rituximab. These two trials, however, resulted in conflicting findings, where addition of rituximab led to improved outcomes in IELSG 32, but no significant difference in HOVON105/ALLG NHL24. Using these two trials, Schmidt et al. conducted a trial-level random-effects meta-analysis to determine whether addition of rituximab would impact OS and PFS. They found that OS was not significantly improved as determined by a hazard rate (HR) of death in the pooled analysis at 0.76 [95% CI, 0.52-1.12], but reported rituximab may improve PFS with HR for PFS at 0.65 [95% CI, 0.45-0.95], albeit with low certainty of evidence [49]. Similarly, Fritsch et al. reported a single center prospective phase II study in 28 elderly patients (age ≥ 65) and found that the addition of rituximab to HDMTX + lomustine + procarbazine (R-MCP) improved PFS but not OS compared to MCP [50].

Limitations
A limitation of this review is the lack of patient-level data which precluded not only statistical comparison among the cohort categories but also multivariate analyses that would have elucidated predictors of outcomes and confounding variables. This lack of patient-level data also limited our reporting of treatment-related toxicity results to the number of cohorts (and not the actual percentage of patients) that reported their occurrences and did not allow for assessment of their impact on outcomes. Similarly, we reported only planned courses of HDMTX induction therapy rather than the actual number of courses because patient-level data were not available. Further, age and performance status are important predictors of prognosis as well as determinants of consolidation therapy modality; the absence of these data in addition to the heterogeneity of modalities made it difficult to assess the impact of consolidation therapy on outcomes. Thus, we were able to only present a descriptive comparison of pooled estimates of ORR and 2-year PFS comparison by number of planned HDMTX courses (<5 vs ≥5), CNS chemotherapy at induction, and provision of RT and SCT without statistical analysis (Table 2). Additionally, there was considerable variability among the treatment cohorts with respect to the administration of leucovorin rescue (specific dose, timing, duration) which would have potentially impacted the frequency of toxicities as well as treatment efficacy (Supplementary Table S2).
The outcomes of interest in this review were ORR (CR + PR) and 2-year PFS. Although OS would have provided a more comprehensive assessment of outcomes for the treatment regimens, ORR and 2-year PFS were selected as the main outcomes in this analysis because they were the most frequently reported in the studies included in this review. ORR and 2year PFS allowed for determination of the immediate effect of HDMTX induction regimens and overall therapeutic efficacy during the first 2 years after diagnosis.
Another limitation is the relative paucity of articles that reported clinical trials in elderly patients, particularly those over the age of 65. Most of the trials in this review were composed of patients with a wide age range varying from 18 to 85, with the median age for the trials ranging from 41 to 63 years; two trials (cohort numbers 253 and 257) had no patients over the 65 years of age (Supplementary Table S1). Only three trials (cohort numbers 67, 77, 82) had patients with a median age ≥ 70 years, and it is notable that no consolidation therapy was given in these trials, which in turn may have impacted PFS [27,31,32]. Although this limitation is likely due to the inherently smaller number of clinical trials of PCNSL that include only elderly patients, it is possible that our results do not properly represent outcomes in this important older age group who are known to have poorer prognoses than younger patients [7,51].

Conclusions
This systematic review summarized prospective clinical trials utilizing HDMTX for the treatment of PCNSL and assessed outcomes with respect to HDMTX dose and combination regimens used for induction therapy. ORR and 2-year PFS were similar for all three HDMTX dose categories (low, <3 g/m 2 ; intermediate 3-4.9 g/m 2 ; high, ≥5 g/m 2 ), and the intermediate dose, specifically 3-4 g/m 2 was most commonly used. HDMTX regimens that included rituximab showed a tendency to have higher ORR and 2-year PFS compared to those that did not include rituximab. These findings add to the preliminary evidence supporting that sufficient doses and cycles of HDMTX with the inclusion of rituximab provide therapeutic efficacy for the treatment of PCNSL. Increased efforts are needed to include elderly patients ≥70 years of age in clinical trials to assess therapeutic safety and efficacy of different HDMTX doses, combination chemotherapy, and consolidation modalities.