CNS bridging radiotherapy achieves rapid cytoreduction before CAR T-cell therapy for aggressive B-cell lymphomas

Primary and secondary central nervous system lymphoma (CNSL) are rare, aggressive manifestations of non-Hodgkin lymphoma (NHL). CD19 chimeric antigen receptor T-cell therapy (CART) has revolutionized the treatment of extracranial NHL with emerging but promising response signals for CNSL.^1-6 However, concerns about the durability of response after CART^7,8 suggest that additional strategies to maximize CNS control are warranted. Moreover, ∼80% of patients receiving CART require bridging therapy because of symptomatic or rapidly progressing disease,^1,4,9 yet few data exist to guide optimal CNS bridging strategy.¹⁰ 

Radiotherapy is a promising CART bridging strategy for CNSL for multiple reasons. First, bridging radiotherapy (BRT) for extracranial lymphoma is safe, has rapid cytoreductive power, and appears to reduce relapse risk in irradiated high-risk sites.^11-14 Second, salvage radiotherapy for recurrent or refractory CNSL is associated with high response rates of 67% to 88%.^15,16 Third, BRT avoids systemic toxicity associated with chemotherapy, which can hinder subsequent receipt of CART.¹⁰ Fourth, BRT may enhance CART efficacy by preventing antigen escape.¹⁷ 

Despite compelling rationale, specific concerns about CNS-directed BRT (CNS-BRT) remain, given the potential for excess neurotoxicity. CART–associated neurotoxicity is driven by endothelial activation, blood-brain barrier disruption, monocyte dysregulation, cytokine release, and tumor inflammation.^18-21 Given that radiotherapy can lead to enhanced permeability of the blood–brain barrier and augment local inflammation,^22,23 CNS-BRT has a theoretical potential to enhance CART neurotoxicity. We report our experience of safety, response, and relapse pattern after CNS-BRT before CART.

We included all consecutively treated patients with pathologically confirmed NHL involving the CNS who received BRT at the Memorial Sloan Kettering Cancer Center in the period spanning 30 days before leukapheresis through commercial CAR T-cell infusion between 2019 and 2023, to ensure a minimum follow-up of 6 months. CNS involvement included disease infiltration of the brain or spinal cord parenchyma or leptomeningeal space. We also included patients with significant epidural involvement for relevant toxicity assessment because the BRT field would encompass the adjacent spinal cord.

CNSL-directed therapy was defined as systemic therapy prescribed for a known diagnosis of CNSL, as opposed to prophylaxis before demonstrated CNS disease. Systemic therapy was defined as concurrent if there was any overlap with BRT, that is, systemic therapy and BRT were administered at least once on the same day. This research was approved by the Memorial Sloan Kettering Cancer institutional review board. All patients provided signed informed consent for treatment with immune effector cells.

CNS-BRT was delivered using intensity modulated radiotherapy or conventional techniques. Whole-brain radiotherapy fields were treated with standard opposed fields to the inferior C2 vertebra. Partial brain radiotherapy fields encompassed a radiographically defined clinical target volume (CTV) with a 3-mm to 10-mm geometric expansion for the planning target volume. For intact tumors, the CTV was defined as radiographically visible disease with up to 5-mm expansion to account for possible microscopic spread. Involved site spine fields generally encompassed a radiographically defined CTV with 1.5- to 3-cm superior and inferior margin to field edge using conventional planning.

Cytokine release syndrome (CRS) and immune effector cell–associated neurotoxicity syndrome (ICANS) were graded by an institutional multidisciplinary conference according to the American Society of Transplantation and Cellular Therapies consensus criteria²⁴ for 11 of 12 patients; 1 patient received CART at an outside institution and toxicity was graded per outside physician report. Immune effector cell encephalopathy (ICE) scores were graded per physician report. Additional CNS-BRT–related toxicities were graded according to Common Terminology Criteria for Adverse Events version 5.0.

CNS response to BRT and subsequent relapse were evaluated for patients with CNSL but not for patients who received CNS-BRT for purely epidural lesions. CNS response of parenchymal and leptomeningeal lesions were evaluated on postcontrast magnetic resonance imaging (MRI). Response evaluations were performed by an expert CART review committee and an independent, experienced blinded neuro-oncology radiologist (J.S.) to achieve clinical consensus. Evaluated lesions were identified on postcontrast MRI on the series with the thinnest slice thickness using the plane in which the lesion had the greatest long axis diameter. Total lesion size was defined as the sum of product diameters for all measurable lesions. The product diameter is the longest axis diameter multiplied by the orthogonal diameter for a single lesion; this value is calculated for each lesion and summed together. For response and relapse evaluation, comparisons were done in the same imaging plane. Spinal intramedullary disease was defined as involvement of soft tissue structures within the thecal sac. Complete response (CR) was defined as disappearance of all contrast enhancing tumor; small residual areas of contrast enhancement could be considered CR if they demonstrated stability on multiple longitudinal scans in the absence of further therapy. Patients with evidence of leptomeningeal disease at the time of BRT must have also had documented negative cerebrospinal fluid (CSF) cytology to be considered in CR. Partial response (PR) was defined as at least a 50% reduction in the size of contrast enhancing tumor, without the development of any new tumors. Progressive disease (PD) was defined as growth of contrast-enhancing tumor by >25% or the development of any new tumors. Stable disease included tumors that did not achieve PR and did not show evidence of PD. CNS relapse was defined as PD within the CNS parenchyma or leptomeninges. Cytoreduction due to BRT was calculated by determining the change in total lesion size at pre-BRT baseline compared with after BRT but before CAR T-cell infusion. Patients were excluded from cytoreduction analysis if they did not undergo interim MRI between the end of CNS-BRT and CAR T-cell infusion.

Patient and treatment characteristics are reported as median and range for continuous variables and counts for categorical variables. The cumulative incidence of CNS relapse measured from CAR T-cell infusion was calculated by cumulative incidence functions assuming death as a competing risk, reported with 95% confidence intervals (CIs). Follow-up time among survivors was estimated using the reverse Kaplan-Meier method. Significance for quantitative cytoreduction was calculated using a paired sample t test, and the association between CNS tumor size by high-grade ICANS (grade ≥3) using an unpaired 2-sided t test. Statistical significance was set at P < .05 and performed using Graphpad Prism (version 9.3.1, GraphPad Software) and R Studio (version 4.1.2, R Foundation for Statistical Computing).²⁵ 

We identified 12 eligible patients who received CNS-BRT before CART (Table 1). Median follow-up among survivors was 11.8 months (interquartile range, 8.5-21.9). Diagnoses included diffuse large B-cell lymphoma (n = 8), mantle cell lymphoma (n = 2), Burkitt lymphoma (n = 1), and primary CNS lymphoma/diffuse large B-cell lymphoma (n = 1). Involved CNS sites included brain parenchyma (n = 6), leptomeninges (n = 4), and spinal epidural space (n = 2). Five patients had concomitant systemic disease at the time of CNS-BRT. Median age at CNS-BRT was 60 years (range, 30-76), and median Karnofsky Performance Status was 80 (range, 50-90). Nine patients were symptomatic from their CNS disease at the time of CNS-BRT. The median number of prior CNSL-directed therapies was 2 (range, 0-4). Nine patients received prior high-dose methotrexate (HD-MTX) with a median time of 46 days between the last HD-MTX infusion and the start of CNS-BRT (range, 0-393). Five patients had received prior autologous hematopoietic stem cell transplantation.

All 10 patients with CNSL (parenchymal and leptomeningeal involvement) had progressive CNS disease at the time of CNS-BRT. Of 2 patients with epidural disease, 1 patient had PD whereas the other had radiographically stable but symptomatic lymphoma. CNS-BRT fields included whole brain (n = 3), involved site “partial” brain (n = 4), involved site spine (n = 4), and orbits (n = 1). The median treatment dose was 22 Gy^15-33 in 10 fractions.^5-12 Seven patients received systemic therapy as part of bridging, in addition to CNS-BRT; systemic therapy bridging included MTX (HD-MTX [n = 4], intrathecal-MTX [n = 1], rituximab [n = 2], ibrutinib [n = 2], cytarabine [n = 1], pembrolizumab [n = 1], and venetoclax [n = 1]). Four patients progressed on systemic therapy bridging before initiation of CNS-BRT. In general, cytotoxic chemotherapy and CNS-BRT were given sequentially. However, 1 patient initiated CNS-BRT concomitantly with the last dose of HD-MTX because of progressive symptoms. Ibrutinib (n = 2) and venetoclax (n = 1) were given concurrently with CNS-BRT.

Median time from apheresis to CAR T-cell infusion was 66 days (range, 25-199). Median time from CNS-BRT completion to CAR T-cell infusion was 20 days (range, 10-212). Ten patients received fludarabine/cyclophosphamide lymphodepleting therapy, and 2 received bendamustine. CART products included lisocabtagene maraleucel (n = 7), tisagenlecleucel (n = 4), and axicabtagene ciloleucel (n = 1).

Eight patients (75%) experienced mild toxicities possibly attributable to CNS-BRT, all of which were grade ≤2. Toxicities that occurred in >1 patient included fatigue (n = 6 grade 1, n = 1 grade 2), dermatitis (n = 2 grade 1), and cognitive disturbance (n = 1 grade 1, n = 1 grade 2). Toxicities that were observed once included grade 1 headache and grade 1 depressed level of consciousness. No toxicities of grade ≥3 were observed.

After CAR T-cell infusion, 8 of 12 patients (75%) experienced CRS; most events were grade ≤2 (grade 1 [n = 3], grade 2 [n = 4]). One patient experienced grade 3 CRS. Five of 12 patients (42%) experienced ICANS, 3 of whom experienced high-grade ICANS (grade 3 [n = 1], grade 4 [n = 2]). Two of the high-grade cases were reversible (Table 1). All patients were on seizure prophylaxis. The 3 patients who developed severe ICANS had impaired neurologic function at baseline, which was likely multifactorial with contributions from CNSL and other factors. Patient 3 had baseline Karnofsky Performance Status of 50 to 60; ICE score of 5; and intermittent confusion, hallucinations, and disorientation to place. After CAR T-cell infusion, she developed severe encephalopathy requiring intubation, however the syndrome was transient and neurologic status recovered to baseline. Patient 7’s baseline ICE score of 6 and neurologic assessment was “awake, but not alert” with intermittent confusion, aphasia, and documented focal seizure. The patient developed grade 4 ICANS with seizures, communicating hydrocephalus, altered mental status, and myelitis that was irreversible. Patient 9 had grade 4 ICANS characterized by seizure activity 16 days after CAR T-cell infusion, which was reversible. This patient had known baseline epilepsy and the seizure she experienced after CART was of the same semiology as her prior seizures and occurred in the setting of an acute viral infection. There was a trend toward larger CNS tumor size in patients who experienced high-grade ICANS, however this difference was not statistically significant (supplemental Figure 1). Overall, 6 patients received both tocilizumab and steroids.

Of 10 efficacy-evaluable patients, 8 had imaging before and after CNS-BRT and were evaluable for quantitative cytoreduction. Median time from the end of CNS-BRT to the radiographic assessment was 12 days (range, 2-55). Of 8 patients, 3 were on steroids at the time of radiographic assessment. Among 8 evaluable patients, CNS-BRT produced a significant reduction in lesion size (P = .014). All 8 patients had at least a PR without any progressing lesions. The mean reduction in lesion size was 74% (95% CI, 62-86; Figure 1A). Both leptomeningeal and brain parenchymal lesions responded well to CNS-BRT (Figure 1B-C).

The best overall response after CART included n = 6 CR, n = 3 PR, and n = 3 PD. Best CNS response after CAR T-cell infusion included n = 8 CR, n = 1 PR, and n = 1 PD (Figure 2A). The best CNS response after CNS-BRT, which was defined as the maximal decrease in lesion size at any point after CNS-BRT or CAR T-cell infusion, included n = 8 CR and n = 2 PR; that is, 1 patient had a PR after CNS-BRT but developed out-of-field CNS recurrent disease before CART and continued to have further CNS progression after CART. The median time to best CNS response was 170 days (range, 31-307). The mean maximal change in lesion size was −94.8% (95% CI, −89.4 to −100.1; Figure 2B).

Among the CNSL cohort (n = 10), there were no in-field CNS recurrences. The 12-month cumulative risk of any CNS relapse was 25% (95% CI, 6-52), which included 3 relapse and/or progression events (Figure 3A). Of note, the 3 relapse/progression events occurred in patients with active or prior leptomeningeal disease. Patient 1 had radiographic leptomeningeal involvement of the V1 nerve and positive CSF cytology at the time of CNS-BRT. This patient received CNS-BRT to a bilateral orbital field to target the radiographic disease, as well as additional systemic bridging therapy (rituximab, cytarabine, and venetoclax); CSF cytology cleared before CART. He experienced an out-of-field recurrence in the brain 1 month after CAR T-cell infusion. Patient 3 had a bulky lumbar leptomeningeal tumor and positive CSF cytology at the time of CNS-BRT. This patient received CNS-BRT to the lumbar lesion with concurrent ibrutinib. After initial response, she developed a multifocal leptomeningeal recurrence, including at the margin of the prior RT field, before CART. She experienced further progression after CAR T-cell. Patient 12 had a history of positive CSF cytology, but this was negative at the time of CNS-BRT. She received CNS-BRT to a progressive right basal ganglia lesion and ultimately had a periventricular recurrence at the margin of the prior radiotherapy field (Figure 3B).

We report a retrospective series of patients who received CNS-BRT for NHL, including 10 patients with CNSL (1 primary CNSL [PCNSL], 9 secondary CNSL [SCNSL]) and 2 patients who received CNS-BRT to epidural targets. This represents a particularly high-risk and difficult-to-manage cohort because all 10 patients with CNSL had CNS disease progressive through standard chemotherapy, and 11 of 12 patients in the cohort had PD at the time of CNS-BRT. Importantly, CNS-BRT achieved significant cytoreduction of all treated lesions, at a median interval of only 12 days from the completion of BRT. This rapid and robust response is critical as it demonstrates that BRT can achieve tumor control and potentially symptom palliation within the often-narrow timeframe of the CART bridging period. This response was maintained or improved after CART for 9 of 10 patients, at least temporarily. Only 1 patient did not respond to CART and exhibited rapidly progressive leptomeningeal disease, which was fatal. The 12-month estimated risk of CNS relapse and/or progression after CART was 25%, in this limited and heterogeneous cohort.

The overall rate of CRS in this study was 67% (8/12) and high-grade CRS was 8% (1/12). The overall rate of ICANS was 42% (5/12) and 25% high-grade (3/12). The overall rate of toxicity reported here is largely similar to what is reported in the literature for patients with CNSL. A series of 10 patients with active SCNSL treated with CART reported that 3 of 10 patient developed high-grade ICANS and this was not associated with radiotherapy.²⁶ In a meta-analysis, Cook et al examined 128 patients who received CART for CNSL and reported composite CRS rates of 70% for PCNSL (13% high grade) and 72% for SCNSL (11% high grade). They report a 53% rate of ICANS for PCNSL (18% high grade) and 48% rate of ICANS for SCNSL (26% high grade).¹ Epperla et al performed a multicenter retrospective analysis of 61 patients who received CART for SCNSL and reported a 44% risk of high grade ICANS; this was also not associated with the receipt of BRT.²⁷ In combination, our early data support the notion that CNS-BRT is not clearly associated with excess neurotoxicity,² although larger series and prospective validation is required. We acknowledge that only 1 patient in our cohort was treated with axicabtagene, whereas ∼40% of patients in the referenced cohorts received axicabtagene^1,27; axicabtagene is known to have greater ICANS risk versus 4-1BB products.^28-31 

Two patients who experienced severe ICANS in our cohort had impaired baseline neurologic function with decreased ICE scores of 5 and 6 before CAR T-cell infusion, which likely predisposed to CART–associated neurotoxicity. Overall tumor burden has previously been shown to correlate with CART–associated neurotoxicity.³² Patients who developed severe ICANS in our cohort had a numerically larger tumor size at baseline and after CNS-BRT, however this difference was not statistically different. A larger cohort will be needed to determine whether the radiomics of active CNS disease at the time of CART is correlated with the neurotoxicity risk or severity.

At present, given the relatively paucity of data, it may be advisable to use CAR T-cell products with favorable toxicity profile, such as the 4-1BB products, when planning to integrate CNS-BRT. If clinically feasible, CNS-BRT should not be given concurrently with cytotoxic chemotherapy given historical concerns for enhanced toxicity. Combined modality bridging that incorporates CNS-BRT and other systemic agents such as targeted therapies is an interesting avenue to explore but awaits data on safety and tolerability.

There is now an emerging body of literature that suggests that BRT for extracranial lymphoma is associated with better response to CART through rapid cytoreduction and reduced relapse of high-risk lesions.^13,14,33 The data here provide early evidence that, similar to that for extracranial sites, BRT can achieve critical cytoreduction for CNS lesions. Consistent with the notion that BRT can reduce relapse at high-risk sites, there were no in-field, local relapses observed in this series.

All 3 relapses that occurred in this study were associated with either active or former leptomeningeal disease. It is likely that leptomeningeal involvement confers greater risk for CNS relapse after BRT, with prior studies showing that leptomeningeal involvement is associated with poorer outcomes for patients with CNSL treated with radiotherapy.¹⁶ This is not surprising because radiotherapy is typically a focal treatment, whereas leptomeningeal disease is a diffuse process that involves the entire neuroaxis. In the future it may be interesting to explore alternative bridging strategies to optimize control of leptomeningeal disease. For example, very low–dose radiation (4 Gy) has been shown to augment CART tumor killing via death-receptor signaling augmentation; this very low–dose radiation is also associated with minimal toxicity in a large randomized phase 3 study.^17,34 Very low–dose bridging radiation to the craniospinal axis may improve tumor control in the leptomeningeal compartment after CART.

Limitations of this study include the small sample size and retrospective nature. This is in large part because of the rare nature of CNSL. As such, the cohort has heterogeneity of lymphoma histology and CAR T-cell product. The study does not contain a comparator arm of patients who did not receive CNS-BRT. It is challenging to find an appropriate comparator group because this is a selected cohort that typically received CNS-BRT out of necessity after progressing on CNS-directed systemic therapy. Despite these limitations the effect of CNS-BRT on tumor response was consistent across all patients examined.

In the future, it will be important to determine the lesion level characteristics that are associated with CNS relapse so that it may be possible to determine which patients stand to benefit from CNS-BRT. These studies should assess whether CNS-BRT alters the immune effector cell–toxicity profile, optimal timing of CNS-BRT, and whether CNS-BRT improves the overall response and durability of CNS–directed CART. Practically speaking, the successful implementation of a combined modality approach requires significant coordination between medical oncology, cellular therapy, and radiation oncology teams. Overall, this study provides early evidence that CNS-BRT can be an important option in the CART bridging toolbox, especially for CNSL for which the options for bridging therapy may otherwise be limited.

This work is supported, in part, by the La Fundación Española de Hematología y Hemoterapia (FEHH)/Gilead 2022 grant (A.L.D.A.), and a grant from the Alfonso Martin Escudero Foundation (M.C.).

This work was supported by the Memorial Sloan Kettering Cancer Center Comedy vs Cancer Grant Program, Connecticut Cancer Foundation, Lacher Fellowship in Lymphoma Radiation Oncology, LLS Translational Research Program grant 6654-22, the Steven A. Greenberg Award in Lymphoma, and the Memorial Sloan Kettering Cancer Center Support grant (P30 CA008748).

The funding sources were not involved in study design, collection, analysis, or interpretation of data, nor in the writing of the report or decision to submit for publication.

Contribution: G.Y.C., J.Y., and B.S.I. designed research; G.Y.C., S.M.D., and B.S.I. analyzed data; G.Y.C., J.S., and C.G. evaluated neuroradiologic response; G.L.S., R.S., P.B.D., R.J.L., M.-A.P., M.L.P., M.S., and C.G. evaluated clinical response and toxicity; G.S., R.S., H.H., K.T., A.A.T., B.F., C.H., A.B., A.L.D.A., M.C., G.C., P.B.D., R.J.L., P.G., G.S., M.-A.P., M.L.P., L.F., M.S., C.G., and J.Y. discussed results and edited the manuscript; and G.Y.C. and B.S.I. wrote the manuscript, had full access to all of the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis.

Conflict-of-interest disclosure: G.L.S. has received research funding to the institution from Janssen, Amgen, Bristol Myers Squibb, Beyond Spring, and GPCR, and is on the data and safety monitoring board for ArcellX. A.B. has received consulting fees from Bristol Myers Squibb. A.L.D.A. reports research funding from Kite/Gilead. M.S. served as a paid consultant for McKinsey & Company, Angiocrine Bioscience, Inc, and Omeros Corporation; received research funding from Angiocrine Bioscience, Inc, Omeros Corporation, and Amgen, Inc; served on ad hoc advisory boards for Kite, a Gilead company; and received honoraria from i3Health, Medscape, and CancerNetwork for Continuing Medical Education-related activity. G.S. has received, in the last 12 months, financial compensations for consulting from AbbVie, ATB Therapeutics, BeiGene, Bristol Myers Squibb, Genentech/Roche, Genmab, Innate Pharma, Incyte, Ipsen, Kite/Gilead, Modex, Molecular Partners, Orna Therapeutics, Treeline; is a shareholder in Owkin; and has received research support managed by his institution from AbbVie, Genentech, Genmab Janssen, Ipsen, and Nurix. M.L.P. reports honorarium and research funding from Bristol Myers Squibb, Cellectar, Ceramedix, Juno, Kite, MustangBio, Garuda Therapeutics, Novartis, Pluto Immunotherapeutics, Rheos, Seres Therapeutics, Smart Immune, Thymofox, and Synthekine, and other support from Juno and Seres. M.-A.P. reports personal fees from Adicet, Allovir, Caribou Biosciences, Celgene, Bristol Myers Squibb, Equilium, Exevir, Karyopharm, Merck, MorphoSys, Omeros, Syncopation, VectivBio AG, Vor Biopharma, Cidara Therapeutics, Medigene, Sellas Life Sciences, and NexImmune; received personal fees and other support from Incyte, Kite/Gilead, Miltenyi Biotec, Nektar Therapeutics, and Novartis; and other support from OrcaBio, outside the submitted work. B.S.I. reports honorarium from GT Medical Technologies, and research support (to the institution) from AstraZeneca, Bayer, GT Medical Technologies, Kazia Therapeutics, and Novartis. The remaining authors declare no competing financial interests.

Correspondence: Brandon S. Imber, Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, 530 E 74th St, New York, NY 10065; email: imberb@mskcc.org.