Outcomes of patients with secondary central nervous system lymphoma treated with CAR T-cell therapy: A CIBMTR analysis

To the Editor,

Secondary central nervous system lymphoma (SCNSL) generally carries a poor prognosis. While the advent of chimeric antigen receptor T-cell therapy (CAR-T) revolutionized the therapeutic landscape of relapsed or refractory (R/R) LBCL^1-5, most CAR-T trials excluded patients with CNS involvement. Furthermore, there are only a limited number of studies that evaluated the impact of CAR-T in SCNSL patients^3,6-15. Hence, we assessed the outcomes of LBCL patients with SCNSL who received CAR-T using the Center for International Blood and Marrow Transplant Research (CIBMTR) registry data.

All adult patients (≥18 years) with LBCL who received CD19-directed commercial CAR-T for SCNSL from June 2015 to March 2022 in the CIBMTR registry were included. Patients who received CAR-T products other than axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), or lisocabtagene maraleucel (liso-cel) were excluded. The primary endpoint was overall survival (OS). Secondary endpoints included evaluation of toxicity, clinical outcomes, and hematopoietic recovery. Toxicity outcomes included cytokine-release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). Clinical outcomes included response rates (overall response rate [ORR], and complete response [CR] rate), progression-free survival (PFS), non-relapse mortality (NRM), and progression/relapse. Table S1 details the definition of endpoints and statistical analysis.

144 patients met the eligibility criteria (see Consort Diagram, Figure S1). Table 1 shows the characteristics of the SCNSL patients who received CAR-T. The median age at CAR-T infusion was 61 years (23-83 years) with 35% of the patients ≥65 years. Sixty-two percent (n=89) of patients had parenchymal involvement, while 28% (n=40) had CSF/leptomeningeal involvement. Ninety-seven percent (n=139) of patients had active disease prior to CAR-T. Among these, 93 patients also had systemic involvement. Ninety-three patients received a prior regimen containing high-dose methotrexate. The median time from apheresis to CAR-T infusion was 33 days. The most frequently used CAR-T product was tisa-cel (n=87, 60%), followed by axi-cel (n=48, 33%), and liso-cel (n=9, 6%). The median follow-up of survivors was 24 months (range, 2-52 months).

Table 1:

Baseline characteristics of SCNSL patients who underwent CAR-T

Variable	All patients, N=144 (%)
Patient related
Age at CAR-T infusion
Median, range (yrs)	61 (23-83)
18-64 yrs	93 (65)
≥65 yrs	51 (35)
Sex
Males	52 (36)
Females	92 (64)
Race
White	113 (79)
Asian	8 (6)
African-American	6 (4)
More than one race	1 (1)
Not reported	16 (11)
Ethnicity
Hispanic or Latino	12 (8)
Non-Hispanic or non-Latino	119 (83)
Unknown	13 (9)
Karnofsky performance score prior to CAR-T
90-100	44 (31)
70-80	68 (47)
<60	17 (12)
Not reported	15 (10)
HCT-CI
0	43 (30)
1	29 (20)
2	16 (11)
≥3	52 (36)
Not reported	4 (3)
Disease related
Site of CNS involvement at any time prior to CAR-T infusion
Parenchymal	89 (62)
CSF/Leptomeningeal	40 (28)
Parenchymal + CSF/Leptomeningeal	15 (10)
Primary refractory disease	60 (42)
Early treatment failure	121 (84)
CAR-T related
Median lines of prior therapies, range	3 (1-9)
Types of prior HCTs
No prior HCT	99 (69)
Prior allo-HCT	1 (1)
Prior auto-HCT	42 (29)
Prior auto and allo-HCT	1 (1)
Not reported	1 (1)
Median time from diagnosis to CAR-T, range (months)	14 (2-290)
Disease status prior to CAR-T infusion
CR	5 (4)
Not CR/Active Disease	139 (97)
Bridging therapy
No	56 (39)
Yes	81 (56)
Not reported	7 (5)
Type of bridging therapy (n=81)
Chemotherapy*	53 (65)
Monoclonal antibodies	10 (12)
BTKi/IMID therapy	5 (6)
Intrathecal/intraocular	7 (9)
Radiation therapy	4 (5)
Others	2 (2)
CAR-T product
Axi-cel	48 (33)
Tisa-cel	87 (60)
Liso-cel	9 (6)
Year of CAR-T
2018	14 (10)
2019	33 (23)
2020	42 (29)
2021	48 (33)
2022	7 (5)
Median follow-up of survivors in months (range)	24 (2-52)

Abbreviations: HCT- Hematopoietic cell transplantation, yrs- years, HCT-CI- HCT comorbidity index, CNS-central nervous system, CSF-cerebrospinal fluid, auto-HCT- autologous HCT, allo-HCT- allogeneic HCT, CR-complete response, BTKi- Bruton tyrosine kinase inhibitor, IMID- immunomodulatory drugs, axi-cel- axicabtagene ciloleucel, tisa-cel- tisagenlecleucel, liso-cel- lisocabtagene maraleucel

^*Includes both single agent chemotherapy (n=33) and multiagent chemotherapy (n=20)

Among the patients who developed CRS, the median time to CRS was 4 days (1-10 days). Any grade CRS developed in 108 (75%) patients with 17 (12%) experiencing grade 3 or higher CRS. The median time to ICANS was 6 days (1-33 days). Any grade ICANS was seen in 51 (36%) patients with 35 (25%) experiencing grade 3 or higher ICANS (Table S2).

Among 136 response evaluable patients, the ORR/CR at day 100 was 68% and 53%, respectively. The 1-month cumulative incidence of neutrophil recovery was 90% (95%CI=85-94), while the 100-day cumulative incidence of platelet recovery was 82% (95%CI=75-88). The 2-year cumulative incidence of NRM and relapse/progression was 5% (95%CI=2-10, Figure 1A) and 74% (95%CI=65-81, Figure 1B), respectively (Table S3). At two years post-infusion, the probability of PFS and OS was 21% (95%CI=15-29, Figure 1C) and 34% (95%CI=25-43, Figure 1D), respectively. There was no difference in PFS or OS based on the timing of CAR-T (2018-2020 versus 2021-2022). Figures S2 and S3 show the forest plot for PFS and OS among the subgroups. Among these, KPS <90% was the only factor associated with inferior OS (hazard ratio 2.23, 95%CI=1.3-3.8, Figure S3). At the most recent follow-up, 85 (59%) patients were reported as deceased. The most common cause of death was relapse or disease progression (n=69, 81%), followed by infection (n=5, 6%), and organ failure (n=4, 4%). CRS was responsible for 1 (1%) death. Table S4 shows the breakdown of causes of death.

Figure 1.

Outcomes of patients with SCNSL patients who received CAR-T A) NRM B) Relapse/progression C) progression-free survival D) overall survival

In this registry analysis (first-to-date), we make several important observations of SCNSL patients who received CAR-T. First, the ORR and CR rates were 68% and 53%, respectively. Second, the rate of grade ≥3 CRS was 12%, while the rate of grade ≥3 ICANS was 24%, however, there were no ICANS related deaths in the study. Third, the cumulative incidence of NRM was low. Fourth, even though the response rates were high, the responses were not durable as evidenced by high rates of progression/relapse and therapy failure. Lastly, we did not identify any factors associated with inferior PFS; however, KPS <90 was associated with inferior OS. Interestingly, our study showed a higher use of tisa-cel compared to axi-cel and liso-cel. While it is difficult to explain the particular reason(s) that led treating physicians to favor using tisa-cel, it is plausible that the prescribing bias was due to lower perceived toxicity with tisa-cel compared to other products.

We found the ORR was high in SCNSL patients who received CAR-T, which is in line with the previously published studies (Table S5) ^3,6-15. The CR rate in our study was comparable to the recently published study¹⁴ (where the CR rate was 57%) but higher than the other SCNSL studies (which was around 40%)^6,10. One potential reason could be that the contributing centers may not necessarily report the responses using uniform criteria leading to possible over-estimation of CR rates. Despite high CR rates, the responses were not durable as evidenced by dismal PFS. We did not identify any factors prognostic for PFS or a subset of SCNSL patients that may have durable remission with CAR-T.

We found the rate of any grade CRS and grade ≥ 3 CRS was similar to previously published studies^10,14. While the rate of grade ≥ 3 ICANS was comparable to that reported in one study (Karschnia, et al¹⁰, n=10), it was significantly lower relative to the other study (Epperla et al¹⁴, n=61). The CAR-T product may have influenced the rates of ICANS given the known differential risk of neurologic toxicity between CAR-T products using a CD28 versus 4-1-BB costimulatory molecules. In the study by Epperla et al¹⁴, axi-cel was the most commonly used CAR-T product (n=30, 49%)¹⁴ in contrast to tisa-cel in the current study (n=87, 60%). Other potential reasons include the use of radiation therapy as bridging, where 14 patients (23%) received RT+/− chemotherapy in the Epperla et al study¹⁴, in contrast to 4 patients (3%) in the current study.

This study has limitations inherent to observational studies owing to the preferences of treating centers/physicians including the selection of bridging therapy and CAR-T product. It is important to note that data reported in the study is not an intention-to-treat analysis but rather an analysis of patients who had received infused cells. Due to the use of diverse bridging modalities, we could not study the impact of type of bridging therapy on post-CAR-T outcomes. Nevertheless, the administration of bridging did not appear to affect the PFS or OS in the current study.

In conclusion, this is the largest study reporting on outcomes of patients with SCNSL treated with commercial CD19-directed CAR-T. We show that CAR-T is feasible and associated with similar toxicities compared to patients with systemic LBCL and could be considered in eligible patients with relapsed SCNSL. Future studies should evaluate therapeutic strategies post-CAR-T to improve both response rates and durability of remission.

DATA AVAILABILITY STATEMENT

Data will be available on the CIBMTR public website