Prevention and management of secondary central nervous system lymphoma

Abstract

Secondary central nervous system (CNS) lymphoma (SCNSL) is defined by the involvement of the CNS, either at the time of initial diagnosis of systemic lymphoma or in the setting of relapse, and can be either isolated or with synchronous systemic disease. The risk of CNS involvement in patients with diffuse large B-cell lymphoma is approximately 5%; however, certain clinical and biological features have been associated with a risk of up to 15%. There has been growing interest in improving the definition of patients at increased risk of CNS relapse, as well as identifying effective prophylactic strategies to prevent it. SCNSL often occurs within months of the initial diagnosis of lymphoma, suggesting the presence of occult disease at diagnosis in many cases. The differing presentations of SCNSL create the therapeutic challenge of controlling both the systemic disease and the CNS disease, which uniquely requires agents that penetrate the blood-brain barrier. Outcomes are generally poor with a median overall survival of approximately 6 months in retrospective series, particularly in those patients presenting with SCNSL after prior therapy. Prospective studies of intensive chemotherapy regimens containing high-dose methotrexate, followed by hematopoietic stem cell transplantation have shown the most favorable outcomes, especially for patients receiving thiotepa-based conditioning regimens. However, a proportion of patients will not respond to induction therapies or will subsequently relapse, indicating the need for more effective treatment strategies. In this review we focus on the identification of high-risk patients, prophylactic strategies and recent treatment approaches for SCNSL. The incorporation of novel agents in immunochemotherapy deserves further study in prospective trials.

Introduction

Secondary central nervous system (CNS) lymphoma (SCNSL) is defined by the involvement of the CNS, either at the time of initial diagnosis of systemic lymphoma or in the setting of relapse, and can be either isolated or with synchronous systemic disease.¹ The risk of CNS involvement in patients with diffuse large B-cell lymphoma (DLBCL) is approximately 5%; however, the presence of certain clinical and biological features has been associated with a risk of up to 15%.² Due to the poor prognosis of SCNSL, there has been growing interest in improving the definition of patients at increased risk of CNS relapse, as well as identifying effective prophylactic strategies to prevent it. In this review we discuss the clinical presentation, the identification of high-risk patients, prophylaxis strategies and recent treatment approaches for SCNSL as well as consider future directions.

MEDLINE, EMBASE and PubMED were systemically searched for publications in English using the following terms: ‘CNS’ and ‘lymphoma’, ‘secondary CNS lymphoma’. References from relevant publications were also searched.

Clinical presentation

DLBCL may involve the brain, meninges, cranial nerves, eyes, and/or spinal cord, which are considered immune-privileged sites with blood-brain and blood-retinal barriers creating therapeutic challenges. Approximately 40% of patients present with de novo disease and 60% at relapse, either with isolated CNS disease or synchronous systemic involvement.^3,4 Patients who relapse after prior treatment typically do so within 6-9 months,⁵ which may be a consequence of occult CNS malignant cells at diagnosis or a failure of systemic therapy, CNS therapy, or both.

Although historic reports suggested a high proportion of leptomeningeal involvement,⁶ more recent data indicate parenchymal involvement in 40%-60% of patients, leptomeningeal involvement in 20%-30%, and both in 10%.^3,4,7,8 Direct infiltration of tumor cells from craniofacial or epidural masses into the CNS may also occur. Systemic sites of disease are typically both nodal and extranodal.³

Clinical symptoms are often the first indication of CNS disease and may be diverse, reflecting involvement of the CNS as well as, rarely, the peripheral nervous system. Common symptoms include motor deficits, headaches, cognitive impairment, cranial nerve involvement and neuropsychiatric changes^7,9 and, less frequently, blurred vision and floaters in those with ocular involvement. In older patients, CNS relapse may present with more subtle symptoms of asthenia, hearing impairment and urinary incontinence.¹⁰

Diagnosis

Biopsy and staging investigations are ideally performed prior to steroid administration, in order to maximize diagnostic yield, since corticosteroids have been shown to prevent or delay diagnosis in 50% of cases.¹¹ Our suggested diagnostic and staging investigations are outlined in Table 1.

Biopsy

The gold standard for SCNSL diagnosis has been the histopathological analysis of a stereotactic biopsy of the brain or cytological examination of cerebrospinal fluid (CSF). Less commonly, the diagnosis can be achieved by cytological examination of vitrectomy samples.¹² Histological features of these highly cellular, diffusely growing tumors include atypical medium to large cells with pleomorphic nuclei and distinct nucleoli. Malignant cells express pan B-cell antigens (CD19, CD20, CD22, CD79a) with light chain restriction, negative plasma cell markers and a high Ki67 (MIB1) proliferation index. CSF examination includes biochemical analysis, cell count, morphology, flow cytometry and molecular testing. Increased protein concentration may indicate disruption of the blood-brain barrier, often associated with parenchymal lesions, whereas decreased glucose concentration is usually associated with CSF or meningeal infiltration, especially in cases with high tumor lymphocyte counts. In selected cases, in which findings are inconclusive, analysis of tissue or CSF samples for immunoglobulin gene rearrangements may establish B-cell clonality, supporting the diagnosis.¹³ Other tests may improve diagnostic rates and are increasingly used in patients with disease that cannot be biopsied. Assessment of the MYD88^L265P mutation and interleukin-10 levels in the CSF have shown high diagnostic sensitivity and specificity in patients with primary CNS lymphoma (PCNSL), with high concordance rates in paired tissue and CSF samples, independently of the site and burden of disease.¹⁴ The sensitivity and specificity of these and other promising diagnostic tools should be assessed in patients with SCNSL and prospective studies to validate the efficacy of CSF molecular studies are ongoing (NCT05036564). For intraocular investigation, the diagnostic yield is superior with vitrectomy than with core vitreous sampling.¹⁵

Patients with lesions that cannot be biopsied represent a challenge and should, as a standard, be reviewed in a multidisciplinary team setting. Our consensus is that when patients present with concurrent CNS and systemic lymphoma, the diagnosis can be made from a systemic-site biopsy alone if magnetic resonance imaging (MRI) findings are consistent with lymphoma after review by an expert neuroradiologist. Isolated SCNSL may be diagnosed with characteristic brain MRI features alone in the setting of early relapse (i.e., <2 years from initial diagnosis). Biopsy of isolated CNS lesions presenting more than 2 years after the diagnosis of DLBCL is recommended. Decisions should be made in consensus with expert hematologists and neuro-radiologists to exclude all other potential differential diagnoses.

Imaging

Imaging should include both the CNS and systemic compartments. Contrast-enhanced MRI of the brain and spinal cord cannot reliably differentiate histological entities, nor exclude CNS involvement, particularly after the use of steroids. MRI scanning according to the International PCNSL Collaborative Group (IPCG)¹⁶ is recommended, but experience focused exclusively on SCNSL has not been reported. Ideally MRI should be performed prior to lumbar puncture to exclude focal mass effects and/or obstructive hydrocephalus and avoid non-specific meningeal enhancement that occurs after CSF sampling. Expert neuroradiology review is essential as evolving white matter changes may be due to chemotherapy, radiation or aging.

Whole body positron emission tomography (PET) – computed tomography (CT) is recommended to stage systemic disease. Testicular ultrasound is recommended to exclude testicular involvement and ocular assessment to determine any vitreo-retinal involvement, especially if there are visual symptoms.

Identification of patients with a high risk of central nervous system disease

Our approach to CNS prophylaxis is summarized in Figure 1.

Clinical risk factors

The CNS prognostic model (CNS-IPI), including the five standard International Prognostic Index factors (age >60 years, stage III/IV, ≥2 extranodal sites, elevated lactate dehydrogenase and performance status ≥2) and kidney or adrenal gland involvement, stratifies patients into three categories: low (0-2 risk factors), intermediate (2-3 risk factors) and high risk (4-6 risk factors) with 2-year rates of CNS relapse of 0.6%, 3.4% and 10.2%, respectively.² This is a robust model, but it underestimates the risk of CNS relapse of specific extranodal lymphomas associated with a high risk of CNS recurrence (i.e. testicular, breast)^17,18 that usually present with limited-stage disease, and therefore fall into the low or intermediate categories. The risk of CNS relapse following disease in other extranodal sites, such as the uterus, bone marrow or epidural space, is controversial¹⁹ and craniofacial structures may no longer be high-risk sites since the introduction of rituximab.²⁰ The involvement of ≥3 extranodal sites determined by PET/CT was also shown to confer a high risk of CNS relapse in a retrospective analysis of 1,532 patients, with a 3-year cumulative risk of CNS relapse of 15% compared to 2.6% among patients with ≤2 extranodal sites of disease.²¹ The CNS-IPI model does not include biological risk factors recently associated with higher risk of CNS relapse.

Table 1.Diagnostic and staging investigations.

Figure 1.Algorithm for central nervous system prophylaxis. DLBCL: diffuse large B-cell lymphoma; CNS-IPI: CNS International Prognostic Index; CNS: central nervous system; MRI: magnetic resonance imaging; CSF: cerebrospinal fluid; PET: positron emission tomography; CT: computed tomography; SCNSL: secondary central nervous system lymphoma; CMR: complete molecular response; HD-MTX: high-dose methotrexate; SD: stable disease; PR: partial response; PD: progressive disease; IT: intrathecal. *In testicular DLBCL, consider additional intrathecal therapy.

Biological risk factors

Historically, the presence of a MYC translocation along with a BCL2 and/or BCL6 translocation (high-grade B-cell lymphoma with a “double hit” [DHL] or “triple hit” [THL]) has been associated with an increased risk of CNS relapse of up to 50%; however the series yielding these data may have been subject to selection bias since fluorescence in situ hybridization studies were not routinely performed.²² More recent retrospective series showed lower CNS relapse rates of 5-20%.²³ A retrospective analysis of 40 patients with early-stage DHL/THL showed a very low rate of CNS events (n=1), suggesting that other clinical features may play a role in CNS relapse.²⁴

An activated B-cell phenotype, as determined by gene expression profiling, constitutes an independent risk factor for CNS relapse according to recent studies, with a CNS relapse risk of 7-9%.^23,25 A post-hoc analysis of the GOYA trial showed that an activated B-cell subtype, determined by gene expression profiling, together with high-risk CNS-IPI, was associated with a 2-year CNS relapse rate of 15%.²⁵ Two recent studies have utilized multiplatform analyses encompassing point mutations, structural variants and copy-number alterations to define new molecular sub-groups or clusters of large B-cell lymphomas.^26,27 The MCD and C5 clusters include almost exclusively activated B-cell subtypes with a high frequency of MYD88^L265P, CD79, PIM1, and ETV6 mutations. Interestingly, the genetic features of these subtypes overlap with those observed in primary extranodal lymphomas of immune-privileged sites such as PCNSL and testicular lymphoma. Moreover, a recent study of 26 patients with DLBCL who experienced either isolated CNS relapse (n=13) or systemic (non-CNS) relapse (n=13), showed a higher prevalence of the MCD subtype in patients with CNS relapse compared to those with systemic (non-CNS) recurrence (38% vs. 8%).²⁸ Although molecular analysis may identify patients with a high risk of CNS relapse more precisely, further studies are required to clarify how this can be incorporated into routine clinical practice.

Baseline screening

Baseline brain imaging and CSF analysis may identify asymptomatic patients with CNS involvement and these patients may benefit from CNS-directed therapies. Cytology is a highly specific test with very limited sensitivity, whereas flow cytometry is a more sensitive tool to detect occult CNS disease.²⁹ In a multicenter study analyzing pre-treatment CSF samples from high-risk DLBCL (n=246) and Burkitt lymphoma (n=80), flow cytometry detected CNS disease in 13% of DLBCL and 11% of Burkitt lymphoma patients whereas cytology was positive in only 4% and 6% of cases, respectively.²⁹

Increased levels of soluble CD19 protein in the CSF were associated with parenchymal CNS lymphoma in a multicenter study including 91 patients with high-risk DLBCL.³⁰ The potential role of CSF circulating tumor DNA to predict CNS relapse in patients with systemic B-cell lymphoma with a high risk of CNS relapse was first explored in a study analyzing tumor mutations in CSF samples from 12 patients with B-cell lymphoma collected at diagnosis and during frontline treatment.³¹ CSF analysis detected MYD88 and ASXL2 mutations in one of two patients who relapsed in the CNS in a CSF sample collected 3 months prior to the relapse. No mutations were found in the CSF samples from patients without CNS relapse. More recently, a second study identified clonotypic DNA in the CSF from eight of 22 patients with newly diagnosed B-cell lymphoma; two of the eight with positive CSF circulating tumor DNA eventually relapsed in the CNS, resulting in a 12-month cumulative incidence of CNS relapse of 29%.³² Further studies including a larger number of patients are warranted to explore the potential utility of CSF circulating tumor DNA in identifying patients at higher risk of CNS events.

Strategies for prophylaxis of central nervous system disease

Intrathecal chemotherapy

Prophylaxis with intrathecal (IT) methotrexate (MTX) and/or cytarabine, often combined with steroids, has been used historically in aggressive B-cell lymphomas.³³ However, in the rituximab era, the majority of retrospective studies and post-hoc analyses from prospective trials showed lack of efficacy of IT prophylaxis (Table 2).³⁴ Recent retrospective series including older patients and high-risk DLBCL have shown similar results with no apparent benefit of IT prophylaxis.^5,34-36

Testicular DLBCL represents a particular scenario in which IT prophylaxis might have a role in the prevention of CNS disease according to data from two prospective single-arm studies conducted by the International Extranodal Lymphoma Study Group (IELSG). The IELSG10 study (n=53) showed a low risk of CNS relapse for patients treated with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone) plus contralateral testicular irradiation and four doses of IT MTX (5-year cumulative risk of 6%) compared to patients in previous retrospective series.³⁷ Moreover, after a median follow-up of 5 years, no CNS relapses occurred in the IELSG30 trial analyzing 54 patients treated with R-CHOP, contralateral radiotherapy and intensified CNS prophylaxis with two doses of end-of-treatment high-dose (HD)-MTX (1.5 g/m²) plus four doses of IT liposomal cytarabine.³⁸ These trials have informed clinical practice and as a result many centers have incorporated IT MTX and end-of-treatment HD-MTX as CNS prophylaxis in this particular lymphoma.

Table 2.Studies with more than 400 patients evaluating the use of intrathecal prophylaxis.

High-dose methotrexate

Over recent years, HD-MTX (≥3 g/m²) has been proposed as a potentially better prophylactic strategy in patients with high-risk DLBCL since the majority of relapses in the rituximab era occur in the brain parenchyma. Initial retrospective series suggested a potential benefit of HD-MTX in the prevention of CNS disease; however, in recent years, several large retrospective studies have failed to demonstrate a reduction in CNS relapse (Table 3). A recent multicenter study including 906 patients, of whom 326 were at high risk, showed a CNS relapse risk of 12.2% for patients receiving HD-MTX compared with 11.2% for patients with no prophylaxis.³⁹ Orellana-Noia et al. suggested no benefit of HD-MTX over IT MTX in a series of 1,162 patients from 21 US academic institutions who received CNS prophylaxis (IT MTX n=894, HD-MTX=236), with a CNS relapse rate of 5.4% versus 6.8%, respectively.⁴⁰ Preliminary results from the largest retrospective series published, including 2,300 high-risk patients, also documented a lack of efficacy of HD-MTX with a 5-year incidence of CNS relapse of 9.1% for patients who received HD-MTX versus 8.4% for those who did not.⁴¹ A major limitation of these retrospective reports is that the definition of patients with a high risk of CNS relapse differs greatly between the studies, and the distribution of risk subgroups (i.e., involvement of extranodal sites) varies between the subgroups compared. Patients frequently receive variable numbers of HD-MTX cycles, with or without IT MTX. Finally, there is likely treatment selection bias since younger patients with good performance status are usually more likely to receive CNS prophylaxis than older or unfit patients.

There has been no consensus on the optimal dose or timing of HD-MTX. Wilson et al. conducted a multicenter retrospective study of 1,384 patients treated with R-CHOP-like regimens and HD-MTX prophylaxis, either intercalated or at the end of treatment, and concluded that there was no difference in CNS relapse risk between patients treated with either of the two strategies.⁵ Furthermore, intercalated HDMTX was associated with increased toxicity resulting in a delay of subsequent R-CHOP in 19.3% of patients. These results suggest that, when administrated, HD-MTX should be given at the end of R-CHOP treatment.

Incorporation of novel agents for central nervous system prophylaxis

Small molecules such as lenalidomide and ibrutinib have demonstrated activity as single agents in relapsed/refractory PCNSL,^42,43 and their good CNS bioavailability suggests that they could play a role in preventing CNS relapse when used in combination with R-CHOP. The addition of lenali-domide to R-CHOP in DLBCL showed a lower than expected rate of CNS relapse in a retrospective analysis of 136 patients from phase II trials, with a 2-year CNS relapse rate of 5% in high-risk patients.⁴⁴ However, a recent post-hoc analysis of the phase III trial REMARC reported that maintenance with lenalidomide after R-CHOP in older patients (60-80 years) was not associated with lower CNS recurrence rates.⁴⁵ The two randomized trials evaluating R-CHOP versus lenalidomide plus R-CHOP have not reported CNS-specific outcomes yet.^46,47 The PHOENIX phase III trial comparing R-CHOP plus ibrutinib versus R-CHOP in activated B-cell DLBCL showed CNS relapse rates of 2.4% versus 3.8%, respectively.⁴⁸ The POLARIX phase III study comparing R-CHOP versus R-CHP (rituximab, cyclophosphamide, doxorubicin and polatuzumab, an antibody-drug conjugate targeting CD79b) in intermediate/high-risk DLBCL found similar CNS event rates in both treatment groups (2.7% and 3%, respectively).⁴⁹ Specific clinical trials focusing on high-risk patients including the new molecular classification are essential to evaluate the potential activity of these and other novel therapies in the prevention of CNS relapses.

Prognosis of secondary central nervous system lymphoma

Analysis of real-world, retrospective data from 173 patients treated with varyingly intensive chemotherapy regimens with curative intent identified patient-related factors of age (>60 years), performance status (>1) at SCNSL diagnosis, as well as disease-related factors of combined parenchymal and leptomeningeal involvement (vs. either alone), and SCNSL development during front-line therapy as adverse prognosticators for overall survival on multivariate analysis.⁵⁰ Treatment-related factors, including an adequate dose of MTX to penetrate the CNS, are also important. On univariate analysis of 44 patients with treatment-naïve (de novo) SCNSL treated with mainly R-CHOP-like therapy and HD-MTX, MTX dose (3.5 g/m² vs. lower doses) in induction predicted progression-free and overall survival.⁵¹ Response to induction therapy, employing different regimens, is also prognostic according to retrospective studies.⁵¹ In the largest prospective trial, the mode of presentation (treatment-naïve vs. relapsed) and complete response to frontline chemotherapy (MATRix: rituximab, methotrexate, cytarabine, thiotepa) were independently significant predictors for progression-free survival.⁴

Treatment approach for secondary central nervous system lymphoma

There is a lack of randomized trial data to compare regimens, no international consensus guidelines and consequently wide variation in clinical practice. The majority of our suggested treatment recommendations are based on phase II studies, retrospective series and expert consensus. The most important guiding principles are assessment of patients’ fitness and frailty, duration of initial response to prior therapy, the use of a class of agents to which the patient has not previously been exposed and the burden of present disease/mode of presentation. As a standard, enrolment in clinical trials is encouraged at all stages of the treatment pathway in this rare disease. We outline our suggested approach in Figure 2.

Table 3.Larger retrospective studies evaluating the use of high-dose methotrexate as central nervous system prophylaxis.

Treatment-naïve secondary central nervous system lymphoma (de novo presentation)

The MARIETTA single-arm phase II trial is the largest prospective study conducted so far in patients with SCNSL (Table 4).⁴ The study included patients aged 18-70 years with all modes of presentation: de novo (n=32), relapsed concomitant SCNSL (n=28) and relapsed isolated SCNSL (n=15). Patients received three courses of MATRix followed by three courses of RICE (rituximab, ifosfamide, carboplatin and etoposide), with IT therapy and carmustinethiotepa conditioned autologous stem cell transplantation (ASCT) consolidation. One or two courses of R-CHOP were allowed as initial therapy in patients presenting de novo who had extensive or bulky systemic disease during the first weeks after diagnosis. Patients with de novo presentation achieved the best outcomes with an overall response rate after immunochemotherapy of 75% (complete response rate of 55%), and a 2-year progression-free survival of 71%.

The SCNSL1 study evaluated the combination of HD-MTX and cytarabine followed by R-HDS (cyclophosphamide, cytarabine and etoposide) and carmustine-thiotepa conditioned ASCT in 38 patients (18-70 years) of whom 14 (42%) had treatment-naïve DLBCL.³ In the latter subgroup, ten patients (71%) achieved a complete response with 2-year event-free and overall survival rates of 48% and 41%, respectively (unpublished data). Two patients died because of toxicity.

Dose-intensive regimens represent an alternative option for young and fit patients. A phase II trial of 111 patients with newly diagnosed high-risk DLBCL, including ten with treatment-naïve SCNSL treated with R-CODOX-M/R-IVAC (rituximab, cyclophosphamide, vincristine, doxorubicin and HD-MTX alternating with ifosfamide, etoposide and HD-cytarabine) reported a 2-year progression-free survival of 70% in the SCNSL cohort. Of note, in the whole cohort, patients >50 years and those with poor performance status tolerated treatment poorly and had a 2-year progression-free survival of 43%.⁵²

The combination of R-CHOP plus HD-MTX has also been explored in retrospective series. A collaborative study of the Australasian Lymphoma Alliance analyzed 80 patients with treatment-naïve DLBCL treated with different regimens. Outcomes were similar for patients treated with intensive regimens (HyperCVAD [cyclophosphamide, vincristine, doxorubicin, dexamethasone, and methotrexate with cytarabine] and CODOX-M/IVAC) and R-CHOP plus HD-MTX with 2-year overall survival rates of 55% versus 53%, respectively.⁵³ A small, multicenter study of 41 patients treated mainly with R-CHOP and HD-MTX showed similar outcomes with a 3-year overall survival of 56%.⁵¹ Preferred treatment options for patients with a de novo presentation are outlined in Figure 2.

Figure 2.Treatment algorithm for patients with secondary central nervous system lymphoma. SCNSL: secondary central nervous system lymphoma; CNS: central nervous system; MATRix: methotrexate, cytarabine, thiotepa, and rituximab; RICE: rituximab, ifosfamide, carboplatin and etoposide; R-CODOX-M: rituximab, cyclophosphamide, vincristine, doxorubicin, methotrexate; R-IVAC: rituximab, ifosfamide, etoposide, and high-dose cytarabine; R-CHOP: rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone; IV: intravenous; IT: intrathecal; MTX: methotrexate; BSC: best supportive care; WBRT: whole brain radiotherapy; R-DHAP: rituximab, cytarabine, cisplatin and dexamethasone: MRI: magnetic resonance imaging; PET: positron emission tomography; CT: computed tomography; PR: partial remission; CR: complete remission; ASCT: autologous stem cell transplantation; BTKi: BTK inhibitors; IMID: immunomodulatory drugs; CAR-T: chimeric antigen receptor T cells. *Patients may have one or two cycles of prior R-CHOP as debulking. ** Including IT chemotherapy. Modifications according to age and performance status. ***Novel therapies (including BTKi, IMID, CAR-T) are best in clinical trials.

Table 4.Prospective trials in secondary central nervous system lymphoma.

Relapsed isolated secondary central nervous system lymphoma

CNS-directed approaches for SCNSL have been adapted from those used for PCNSL, and although overall outcomes appear to be inferior in patients with SCNSL, the numbers in prospective series are small (see Table 4).

For patients who are fit, intensive therapy should be offered as outcomes in this setting appear to be comparable to those of patients with treatment-naïve SCNSL. The MARIETTA regimen remains a potential treatment regimen with the most robust prospective trial data. However, MATRix induction alone, with consolidation carmustinethiotepa ASCT, may be a reasonable strategy as the disease is only in the CNS compartment and the overall response rate was 67% after two cycles of MATRix in MARIETTA⁴ and this strategy has been adopted in retrospective series. Dose modification, especially by reducing doses of cytarabine, is commonly employed if patients have impaired performance status or subsequently develop infectious toxicity, and is recommended to reduce morbidity.

For patients not able to tolerate three CNS-directed agents, HD-MTX/cytarabine/rituximab combinations may be an option, particularly for patients >70 years old. The addition of cytarabine to HD-MTX-based regimens improved outcomes in a retrospective review of 80 patients with treatment-naïve SCNSL (2-year overall survival 54% vs. 44%, P=0.037),⁵³ and among 161 patients with isolated SCNSL, there was a trend towards superior outcomes with multi-agent CNS treatment compared with singleagent HD-MTX (P=0.091).⁵⁰ Preferred treatment options for patients presenting with isolated relapse are outlined in Figure 2.

Relapsed concomitant secondary central nervous system lymphoma

These patients have the poorest outcomes in the SCNSL setting.⁵⁰ MARIETTA documented an overall response rate of 46% and 2-year progression-free survival of 14% for 28 patients with synchronous relapse, which appears lower than that in randomized studies of salvage chemotherapy regimens in DLBCL in which 2-year progression-free survival rates were 24-26%.⁵⁵ Most patients relapse early and are therefore resistant to primary therapy in both compartments. The minority are chemo-responsive, but those who undergo ASCT have better outcomes (3-year progression-free survival 40%)⁵⁶ so this should be the treatment goal. Systemic treatment options include RICE and RDHAP (rituximab, cytarabine, cisplatin and dexamethasone) (Figure 2).

Table 5.Retrospective series of patients (N>20) with secondary central nervous system Lymphoma treated with autologous stem cell transplantation.

Patients who are refractory to primary chemotherapy may be candidates for investigational therapeutic approaches including chimeric antigen receptor T-cell therapy (see below). For less fit patients, if the initial response to primary therapy was complete and prolonged, re-treatment with MTX-based chemotherapy may be appropriate, although evidence is sparse in SCNSL. Preferred treatment options for patients presenting with synchronous relapse are outlined in Figure 2.

Role of autologous stem cell transplantation

In SCNSL there are a few non-comparative prospective and retrospective studies showing that consolidation ASCT in first remission is safe and effective and associated with durable responses (Tables 4 and 5). Compared with whole-brain radiation therapy (WBRT) there is reduced neurotoxicity in the long-term in patients with PCNSL.⁵⁷ A dynamic review of a patient’s performance status and overall fitness is recommended to assess transplant eligibility accurately as this may improve significantly after treatment initiation. Four phase II prospective trials support this approach in both treatment-naïve SCNSL and relapsed presentations (Table 4). In these trials, 42-80% proceeded to ASCT. The transplantation rate for salvage chemotherapy regimens in randomized studies of systemic DLBCL were 33-55% in the CORAL,⁵⁸ LY.1 2 ⁵⁹ and ORCHARRD⁵⁵ studies. MARIETTA included the largest number of patients proceeding to ASCT (n=37) and in this study the 2-year progression-free survival was 83%.⁴ Survival benefit was demonstrated in a retrospective review of 60 patients with treatment-naïve SCNSL who were or were not give consolidation with intensive chemotherapy and ASCT: the 3-year progression-free survival rates were 75% vs. 26%, respectively (P=0.001) and the 3-year overall survival rates were 75% vs. 29%, respectively (P=0.002).⁶⁰ ASCT is now increasingly considered a standard of care,⁶¹ with the best outcomes reported in those with treatment-naïve SCNSL and isolated relapse presentations. Unlike PCNSL, there are no randomized trials of ASCT consolidation being compared with another strategy in SCNSL.

Older studies with limited numbers of patients proceeding to ASCT⁵³ or including predominantly BEAM (carmustine, etoposide, cytarabine and melphalan) conditioning have questioned the role of ASCT. However, BEAM has largely been superseded by thiotepa-based conditioning regimens in CNS lymphoma as these latter have superior CNS bioavailability.⁶² In PCNSL, the outcomes following BEAM conditioning are inferior compared with those after thiotepa-based regimens because of higher risk of relapse.⁶³ In another study, the relapse rate with BEAM was 57% at a median of 2.3 months after ASCT,⁶⁴ thus this conditioning regimen has fallen out of favor in CNS lymphoma. A matched cohort of 151 patients with SCNSL undergoing ASCT (of whom 46% had BEAM conditioning) were compared with 4,688 patients without CNS lymphoma and no difference in outcomes was found on matched propensity scoring.⁶⁵

In a retrospective review of 102 patients, multivariate analysis showed that predictors of adverse outcome following ASCT were more than two prior lines of therapy and less than a complete response at ASCT; the 19 patients with both these unfavorable features had a 4-year overall survival of 14%.⁶⁶ Notably, 53% of the cohort received BEAM conditioning, which has now largely been superseded.

The largest series of 134 SCNSL patients undergoing thiotepa-based ASCT reported 3-year overall and progression-free survival rates of 71.6% (95% CI: 61.9%-not reached) and 61.1% (95% CI: 52.2-68.9%), respectively.⁵⁶ One-hundred-day non-relapse mortality was 3% and the cumulative incidence at 1 and 3 years was 8.4% (95% CI: 4.7-14.6). The risk factors determining progression after SCNSL were similar to those prior to ASCT. In multivariable analysis, risk factors for progression-free survival were synchronous relapse presentation (compared with isolated relapse or de novo), age and lines of treatment. Importantly those in partial remission according to MRI or PET-CT prior to ASCT had similar outcomes to those in complete remission, which is consistent with findings in PCNSL. Patients who relapsed after ASCT had poor outcomes and time to relapse after ASCT predicted overall survival.⁵⁶

Allogeneic transplantation is not widely adopted, and there are limited data derived from descriptive series; however, efficacy was described in a case series, albeit with a high 1-year transplant mortality rate of 20%.⁶⁷ The graft-versus-lymphoma effect is thought to be blunted due to the immune privilege of the CNS.

Less intensive consolidation

In a study of 60 patients with treatment-naïve SCNSL barriers to ASCT that were cited included chemorefractory disease, toxicity from induction therapy, age >65 years and physicians’ decisions.⁶⁰ Unsuccessful stem cell harvest is also a factor. It is clear that age itself should not be a restrictive factor in a carefully selected population, with patients up to the age of 70 years in prospective trials⁴ and 77 years in the largest retrospective series receiving thiotepa-based conditioning.⁵⁶ In patients for whom the risk of non-relapse mortality would be considered too high, non-ASCT consolidation strategies have been attempted in small retrospective series; however, outcomes remain limited.

Optimal consolidation for older patients who achieve remission has not been established. WBRT can be effective but neurotoxicity remains a concern. Continuous chemotherapy has been investigated as a consolidation strategy but the follow-up is limited and so this strategy is not routinely recommended. HD-MTX, cytarabine, ifosfamide and liposomal doxorubicin were recently employed in a small retrospective series of 19 patients with SCNSL (10 de novo, 9 relapsed) at a single institute.⁶⁸ At the end of induction, 58% achieved complete remission; the median follow-up was 11 months, progression-free survival was 28 months and overall survival was 34.5 months. Patients in complete remission received consolidation with ifosfamide, etoposide and cytarabine every 3 months, whereas those who did not achieve complete remission were given WBRT. Further data are required to draw conclusions regarding the efficacy of the two strategies.

Role of radiation therapy

WBRT has been found to be effective although responses are usually short-lived, especially when it is used as the sole treatment modality, and relapses outside the radiotherapy field are not uncommon.⁶

WBRT might have a role as consolidation therapy in patients who do not achieve a complete remission after front-line treatment or in those who cannot proceed to ASCT, especially when residual disease is confined to the CNS.^69,70 In the MARIETTA trial, 13 patients received WBRT: seven of the nine patients given WBRT (residual disease, n=5; poor mobilizers, n=2; after ASCT, n=2) after or during immunochemotherapy, to control responsive disease, achieved a complete or partial remission, and only one of them experienced relapse in the CNS; conversely, none of the four patients treated with WBRT for progressive disease responded.⁴

As salvage therapy in SCNSL, earlier retrospective series showed responses to radiotherapy in 67%-88% of patients, including about 50% who achieved a complete remission with a 2-year overall survival of approximately 30%.^71,72 A retrospective study of 44 patients reported that the dominant pattern of relapse after radiotherapy was systemic disease (n=18) and that outcomes were more favorable in patients who received consolidation with ASCT after radiotherapy (n=8).⁷¹

Neurotoxicity is the major long-term complication after WBRT in long-lasting survivors particularly in those >60 years with PCNSL. Importantly, the PRECIS study, conducted in patients <60 years with PCNSL, showed significant neurocognitive decline during follow-up in patients randomized to WBRT consolidation with doses of 40 Gy compared to those randomized to ASCT (64% vs. 13%, P<0.001).⁵⁷ Significant impairments in some attention, memory and execution functions as well as quality of life have been reported in large prospective trials.^70,73 Data from PCNSL have shown that lower doses of radiotherapy, i.e., 23.6 Gy, can be efficacious as a consolidation strategy in patients achieving complete remission after induction, with minimal neurotoxicity, although this approach has been reported predominantly in patients <60 years old.^{74 ,7 5} Although these complications are expected to occur among SCNSL patients as well studies focused on this issue are lacking.

Older/unfit patients

The optimal regimen for treating elderly or frail patients with SCNSL is yet to be defined. Evidence is mainly derived from PCNSL studies. A meta-analysis of 20 PCNSL studies including patients >60 years old found that HDMTX-based therapy was associated with more favorable outcomes than therapies without HD-MTX in elderly patients.⁷⁶

Trials addressing efficacy and tolerability of MATRix in PCNSL and SCNSL have been restricted to patients ≤70 years with Eastern Cooperative Oncology Group performance status ≤2-3. A recent analysis of tolerability and efficacy of MATRix in 156 patients with PCNSL treated in routine clinical practice showed that older and unfit patients (aged >70 years, n=21; with comorbidities, n=13) had a higher risk of infections and worse outcomes than those who would have meet IELSG32 trial inclusion criteria.⁷⁷ A small, prospective study of elderly (69-79 years), fit patients with PCNSL treated with rituximab, HD-MTX and cytarabine followed by busulfan-thiotepa conditioned ASCT showed favorable outcomes in this population.⁷⁸ The addition of rituximab to HD-MTX-based regimens in 38/94 patients with isolated SCNSL was associated with improved overall survival (HR=0.42, 95% CI: 0.25-0.71, P=0.001) with a 44% reduction in risk of death. This was significant even after adjustment for age >60 years, performance status >1, multiagent HD-MTX vs. HD-MTX alone, time to SCNSL and CNS-directed radiotherapy (HR=0.39, 95% CI: 0.22-0.69, P=0.001), and may be considered a less intensive option.⁵⁰ Other combinations for patients with CNS lymphoma who are not eligible for ASCT include rituximab plus HD-MTX and temozolomide⁷⁹ or novel agents.

Other less intensive options for patients considered unfit for MTX-based therapy include corticosteroids, oral chemotherapy with or without rituximab and WBRT for patients with parenchymal disease.⁵⁰ For patients with leptomeningeal involvement, IT chemotherapy alone may be of modest efficacy. Intrathecal MTX, cytarabine, thiotepa and rituximab can be administered into the CSF but need to be given two or three times a week because of rapid clearance. Clarifying the wishes and priorities of the patient is paramount and palliative approaches or best supportive care may be favored in certain situations.

Progression following a secondary central nervous system lymphoma-directed approach

Patients who progress after MTX-based treatment have a dismal prognosis. In the MARIETTA trial, only seven of the 36 (19%) patients who relapsed received salvage therapy, with no responses and a median overall survival after relapse/progression of 1 month.⁴

Novel therapies

Novel therapies show promising preliminary results and are currently under investigation. They have been tested either alone or in combination in patients with CNS lymphoma. The Bruton tyrosine kinase (BTK) inhibitor, ibrutinib, showed encouraging activity in patients with PCNSL which are enriched for MYD88 and CD79 mutations. A phase I study including SCNSL and PCNSL demonstrated that ibrutinib reached therapeutic levels in the CNS and reported clinical responses in five of seven patients with SCNSL, including four complete responses, with a median progression-free survival of 7.4 months.⁸⁰ In a phase II study of 44 patients with relapsed/refractory CNS lymphoma (15 with SCNSL), the overall response rates in patients with SCNSL and PCNSL treated with ibrutinib were 69% and 81%, respectively, with a median progression-free survival of 4 months.⁸¹ Ibrutinib has also been combined with MTX and rituximab with promising results.⁸² An increased risk of aspergillosis has been reported in PCNSL patients treated with combination regimens including ibrutinib and corticosteroids.⁸³ The efficacy of second-generation BTK inhibitors is being investigated in PCNSL patients (NCT04462328). Immunomodulatory drugs, such as lenalidomide and pomalidomide, have also been investigated in relapsed/refractory PCNSL alone or in combination with rituximab with responses, usually of short duration, in approximately 50% of cases.^43,84,85 A recent study of lenalidomide and rituximab in 14 patients with relapsed/refractory CNS lymphoma showed responses in three of eight patients with SCNSL.⁸⁴ The role of lenalidomide as maintenance therapy is being investigated in this setting.

CAR T-cell therapy has shown promising results in patients with CNS lymphoma, with a good safety profile. The TRANSCEND study included six patients with SCNSL of whom three achieved a complete remission, with severe neurological toxicity in two cases.⁸⁶ A small retrospective study also reported complete responses in four of seven patients with SCNSL receiving commercial axicabtagene ciloleucel.⁸⁷ Similarly, another series of eight patients with refractory SCNSL treated with tisagenlecleucel showed a complete response rate of 50% with no significant toxic-ity.⁸⁸ In a phase I/II trial of 12 patients with refractory PCNSL treated with tisagenlecleucel, the overall response rate was 58% and the complete response rate was 50%.⁸⁹ The duration of response to CAR T-cell therapy remains to be defined, as the follow-up of published studies is still short. A number of phase I/II studies are currently evaluating the efficacy of CAR T cells in CNS lymphoma (NCT03484702, NCT04608487, NCT04464200).

Summary of our recommended approach to the management of secondary central nervous system lymphoma

Our recommended approach to the management of SCNSL is illustrated in Figure 2 and summarized here. Participation in prospective clinical trials, especially involving novel agents (BTK inhibitors, immunomodulatory drugs, CAR T cells), is recommended.

With regard to de novo presentation, the preferred options for fit patients include the MARIETTA regimen (MATRix/RICE induction and thiotepa-based conditioned ASCT consolidation in those achieving partial or complete remission or with stable disease prior to ASCT) or R-CODOX-M/IVAC. Less fit patients may achieve responses with rationalized R-MTX-Ara-C/RICE, R-CHOP and intravenous or intrathecal MTX with consideration of ASCT consolidation with thiotepa-based conditioning. Among patients presenting with isolated relapse, for fit patients the preferred options are MATRix induction with carmustine/thiotepa-conditioned ASCT consolidation, or the MARIETTA regimen. Less fit patients may achieve responses with R-MTX-Ara-C-based regimens and ASCT consolidation can be considered.

With regard to patients presenting with synchronous relapse, preferred options for fit patients include MATRix/RICE and ASCT consolidation (MARIETTA approach). Less fit patients may achieve responses with salvage chemotherapy (RICE, RDHAP, etc.) or novel approaches based on time to relapse and availability and with the addition of intravenous or intrathecal MTX at induction and then proceeding to ASCT consolidation (in those achieving partial/complete remission before ASCT). This is an area of unmet need and access to novel approaches, including CAR T-cell therapy and other novel agents, is recommended.

Conclusions

Treatment of SCNSL remains a challenge due to the aggressiveness of the disease, heterogeneity of presentation and the need to address both systemic and brain compartments. The MARIETTA approach has led to long-term responses, especially in patients with treatment-naïve SCNSL; however, outcomes are still dismal in older/unfit patients and in those who relapse after MTX-based treatments. Novel therapies are currently under evaluation, with CAR T-cell treatment showing promising preliminary results in this challenging population.

We need to continue to explore more specific methods of identifying patients at highest risk of CNS relapse, and to investigate more effective prophylactic strategies. Integration of molecular biomarkers with classical clinical risk factors might improve the selection of patients for CNS prophylaxis. Moreover, baseline analysis of CSF circulating tumor DNA may have a role in detecting occult CNS involvement in patients with aggressive B-cell lymphomas who could benefit from CNS-directed therapies. The incorporation of novel agents (immunomodulatory agents, BTK inhibitors) into frontline standard immunochemotherapy might reduce the number of CNS events, although this deserves further study in prospective trials.

Footnotes

• Received August 31, 2022
• Accepted November 9, 2022

Correspondence

K. Cwynarski kate.cwynarski@nhs.net

Disclosures

SB declares speakers’ bureau honoraria from Janssen and Roche and travel support from Gilead. KC declares consulting/advisory fees from Roche, Takeda, Celgene, Atara, Gilead, KITE, Janssen and Incyte; conference/travel support from Roche, Takeda, Kite and Janssen and speakers’ bureau honoraria from Roche, Takeda, Kite and Janssen. AJMF has received speakers’ fees from Gilead and Roche; was a member of advisory boards of Gilead, Juno, Novartis, Pletixa-Pharm, AstraZeneca, BMS, and Roche; currently receives research grants from ADC Therapeutics, Bayer HealthCare Pharmaceuticals, Beigene, Bristol Myers Squibb, Genmab, Gilead, Hutchison Medipharma, Incyte, Janssen Research & Development, MEI Pharma, Novartis, PletixaPharm, Pharmacyclics, Protherics, Roche, and Takeda; and holds patents on NGR-hTNF-a in brain tumors, NGR-hTNF/R-CHOP in relapsed or refractory PCNSL and SNGR-hTNF in brain tumors.

Contributions

All authors contributed equally to writing and editing the paper. All authors reviewed the manuscript and approved its submission.