We performed a multicenter, retrospective, observational study conducted in accordance with the Declaration of Helsinki and approved by the institutional ethics committee of Hospital Universitario Gregorio Marañón.

We included in the CAR-T cohort all adult patients treated with commercially available CAR-T cell products with aggressive B-cell lymphoma different from primary mediastinal large B-cell lymphoma (PMLBCL) who were registered in the GELTAMO/GETH-TC (Grupo Español de Linfomas y Trasplante Autólogo de Médula Ósea/Grupo Español de Trasplante Hematopoyético y Terapia Celular) database of patients treated with CAR T-cell therapy in eight Spanish centers between February 2019 and July 2021. All patients were judged eligible by the Expert Committee of the Spanish National Health System. Patients fulfilling SCHOLAR-1 criteria (9) for refractory disease (progressive disease as best response to any line of therapy, stable disease as best response to ≥4 cycles of first-line therapy or ≥2 cycles of later-line therapy, or relapse <12 months after auto-SCT) were identified and included in the primary analysis. Patient selection, supportive care, toxicity management, and response assessment followed institutional practices. Cytokine release syndrome (CRS) and neurotoxicity were graded according to the according to the ASTCT consensus criteria (23, 24).

This cohort was compared with a historical population of r/r DLBCL patients from the GELTAMO-IPI study, treated in the pre-CAR-T-cell therapy era (pSOC) (25). The GELTAMO-IPI study included patients from 20 academic and community hospitals in the GELTAMO network in Spain, diagnosed between 1998 and 2014, and treated with R-CHOP or more intense regimens with a curative intent. For the present study, centers were required to update the follow-up and additional data about the treatment at relapse or progression. Finally, nine centers (n = 9) agreed to participate in the present study. As previously mentioned for the CAR-T cohort, only patients who met at least one of the Sholar-1 refractory criteria were considered for this cohort as well.

The present analysis was based on a data cutoff on July 25, 2021, and March 1, 2019, for the CAR-T and pSOC cohorts, respectively. Descriptive statistics, including median, and interquartile range (IQR) for the continuous variables and percentages for the categorical variables, were obtained. To evaluate the association between two categorical variables, we employed Fisher’s exact test or the chi-squared test.

The median follow-up time (in months) was calculated for the surviving patients by the reverse Kaplan–Meier method. The time to event, overall survival (OS), and progression-free survival (PFS) were estimated using the Kaplan–Meier method, and comparisons between variables of interest were performed using the log-rank test. The OS for the ITT and infused populations was calculated from start of treatment for the pSOC cohort and apheresis and infusion date, respectively, until the date of death from any cause, censoring for patients alive at last contact. The PFS for the ITT and infused populations was calculated from failure to last treatment and infusion date, respectively, until the date of relapse, progression, or death from any cause, censoring for patients alive and progression-free at last contact. For the comparison analysis between the pSOC and ITT CAR-T cohorts, the survival was measured from the failure to last treatment in the patients fulfilling the Scholar-1 criteria. This analysis was exploratory, and p-values were not corrected for multiple testing. The specific cutoffs for several quantitative variables such as time to approval, apheresis or infusion, ferritin, and C-reactive protein (CRP) were calculated using receiver operating characteristic curves.

To assess the effect of important covariates on response and toxicity, we performed a multivariable logistic regression. We also performed a multivariate survival analysis with the variables that appeared to be significant in the univariate analysis (p < 0.05), according to the Cox proportional hazard regression model. All reported p-values were two-sided, and statistical significance was defined at p < 0.05. Analyses were performed using SPSS version 25 (SPSS, Chicago, IL).

From the initial population of 255 patients registered in the GELTAMO/GETH-TC database, 204 patients meeting the SCHOLAR-1 criteria for refractory disease were included in the intention-to-treat (ITT) analysis. Fifty-one patients were excluded due to non-refractory disease (15 patients), presence of PMLBCL histology (19 patients), histology different from aggressive B-cell lymphoma (4 patients), or lack of data or follow-up (13 patients) ( Supplementary Figure 1 ).

Patient characteristics are shown in Table 1 . Of the 204 refractory patients included in the ITT analysis, 192 (94%) were infused. The median time from official approval to infusion was 60 days (IQR: 49–73 days), and the median time from apheresis to infusion was 46.5 days (IQR: 39–55.7 days). For the infused patients, 101 were administered axi-cel and 91 were administered tisa-cel. Regarding the SCHOLAR-1 criteria in the ITT analysis, 64% of the patients were primary refractory, 82% were refractory to the last therapy, and 27% had an early relapse after auto-SCT. The most common histology included was DLBCL, not otherwise specified (NOS) (77%), followed by transformed follicular lymphoma (13%), high-grade B-cell lymphoma double or triple hit (9%), and high-grade B-cell lymphoma, NOS (1%). Seventeen patients (9%) had an Eastern Cooperative Oncology Group performance status (ECOG-PS) ≥2 pre-lymphodepletion, and 11 (5%) pre-apheresis. The revised IPI (R-IPI) was high (3–5) at diagnosis in 55% of the patients and pre-lymphodepletion in 54% of the patients. Bulky disease (more than 10 cm for the larger diameter) was present in 37% of the patients pre-apheresis. Almost one-third (29%) of the patients previously underwent auto-SCT. Eighty-two percent of the patients required bridging therapy. The most common bridging therapies were gemcitabine-based regimens (35%), cyclophosphamide-based regimens (24%), intense salvage regimens (R-ICE [rituximab plus ifosfamide, carboplatin, and etoposide], R-DHAP [rituximab plus dexamethasone, ara-C, and cisplatin], MINE [mesna, ifosfamide, mitoxantrone, and etoposide]) (13%), bendamustine-based regimens (10%), radiotherapy (5%), steroid monotherapy (5%), lenalidomide-based regimens (2%), nivolumab (1%), and brentuximab (1%).

As shown in Table 1 , characteristics of the pSOC cohort were similar to those of the CAR-T cohort, except for age, which was significantly higher in the pSOC cohort. Similar to the CAR-T cohort, the most common histology included was DLBCL, NOS (84%). The R-IPI was high (3–5) at diagnosis in 58% of the patients. All patients received at least two lines of therapy, with 21% receiving more than three lines previous to the last therapy. Regarding the SCHOLAR-1 criteria, 50% of patients were primary refractory, 71% were refractory to the last therapy, and 17% had an early relapse after auto-SCT. The median number of lines was 2 (IQR: 1–2). The last therapies received were R-GemOx (19%), R-bendamustine (15%), R-ICE (13%), R-ESHAP (etoposide, methyl prednisolone, cytosine arabinoside, cisplatinum) (10%), mini-BEAM (BCNU, etoposide, cytosine arabinoside, melphalan) (5%), R-GIFOX (rituximab plus gemcitabine, ifosfamide, and oxaliplatin) (4%), lenalidomide (4%), and others (30%). One-third of the patients underwent hematopoietic auto-SCT (34%), and 8.6% underwent allogeneic transplantation.

The median follow-up was significantly longer for the pSOC cohort than for the CAR-T cohort ( Table 1 ), 93 months vs. 11 months since start of treatment and apheresis, respectively. Refractory patients treated with CAR T-cells versus pSOC had significantly longer PFS (median of 5.1 and 5.6 vs. 4.6 months, p ≤ 0.001) and better OS (median of 14.5 and 15 vs. 8.2 months, p < 0.001), considering both the ITT and infused populations ( Table 1 ; Figure 1 ). When analyzing separately by the various products versus pSOC, the differences were in favor of both products over the pSOC ( Table 2 ; Figure 2 ), with a median PFS of 4.6 months (95% CI 2.9–8.5) for the pSOC cohort versus 8.5 months (95% CI 6.6–14.4) (p < 0.001) for axi-cel and 4.6 months (95% CI 4.3–8) (p = 0.074) for tisa-cel, and median OS of 8.2 months (95% CI 6.6–9.9) for the pSOC cohort versus not reached (p < 0.001) for axi-cel and 11.7 months (95% CI 7.8–15.6) (p = 0.015) for tisa-cel.

To account for possible imbalances between the groups, we performed a multivariate analysis using Cox regression, including all the patients from both cohorts. We found that pSOC therapy, as well as not having undergone auto-SCT, had a significant adverse impact on PFS and OS ( Supplementary Table 1 ).

Table 3 shows the characteristics of the CAR-T cell cohort at diagnosis and at CAR T-cell therapy. The responses to CAR T-cell therapy are summarized in Supplementary Table 2 . The best response rates after CAR T-cell infusion in this refractory group of patients meeting the Scholar-1 criteria (n = 192) were 36% for complete response, 24% for partial response, 7% for stable disease, and 22% for progressive disease ( Supplementary Table 2 ).

For all infused patients (n = 192), the median follow-up from infusion was 10 months (95% CI: 8–12 months), the median PFS was 3.3 months (95% CI 2–4.6), and the median OS was 11.8 months; the estimated 12-month PFS and OS were 35% (95% CI 27–42) and 50% (95% CI 41–59), respectively, and the 12-month PFS and OS for the patients who achieved complete response were 73% (95% CI 61–85) and 86% (95% CI 76–95), respectively. The factors affecting PFS and OS in the univariate analysis are shown in Table 4 . In the multivariable analysis, the factors with independent adverse influence on PFS were CAR-T cell type (tisa-cel), unfavorable R-IPI at lymphodepletion, non-previous ASCT, and HCTCI (Hematopoietic Cell Transplantation-specific Comorbidity Index) 3–7 pre-lymphodepletion, and those for OS were CAR-T cell type (tisa-cel), unfavorable R-IPI at lymphodepletion, ECOG-PS 2–4 pre-apheresis, primary refractory disease, and HCTCI 3-7 pre-lymphodepletion ( Table 5 ).

For the infused patients (n = 192), axi-cel was superior to tisa-cel for PFS and OS, with a median PFS of 7.3 (95% CI 3.3–11.3) versus 2.8 months (95% CI 2.4–3.1) (p = 0.027), and median OS not reached (NR) versus 10 months (95% CI 6.6–13.4) (p = 0.048), respectively ( Figure 3 ). Similarly, the 12-month PFS was 40% (95% CI 28–52) vs. 28% (95% CI 19–38) (p = 0.023), and the 12-month OS was 58% (95% CI 45–71) and 42% (95% CI 29–54) (p = 0.048) for axi-cel and tisa-cel cohorts, respectively ( Tables 4 , 5 ).

A total of 78% of patients experienced some degree of CRS, with only 6% of cases grade 3 or more. The incidence of CRS was significantly higher in the patients treated with axi-cel than those treated with tisa-cel (90% vs. 65%, p < 0.001); however, the incidence of severe CRS did not differ between the two CAR-T cohorts (8% vs. 4%, p = 0.38). In terms of neurotoxicity, 29% of the patients had some degree of neurotoxicity, with 11% experiencing grade 3 or more. These events were more frequent with axi-cel than with tisa-cel: 42 (42%) vs. 13 (14%) patients, respectively (p < 0.001), and severe neurotoxicity occurred in 16 (16%) and 5 (5%) patients (p = 0.022), respectively. In the multivariable analysis, the factors identified as independently related to CRS were ferritin levels at lymphodepletion >673 UI/l (relative risk [RR] 2.5; 95% CI 1.1–5.6; p = 0.025) and having received axi-cel (RR 4.1; 95% CI 1.8–9.5; p = 0.001). For neurotoxicity, the factors were axi-cel (RR 5.4; 95% CI 2.3–12.5; p < 0.001) and a CRP >173 mg/l at lymphodepletion (RR 3.9; 95% CI 1.8–8.3; p < 0.001).