Impact of bridging therapy on efficacy of CD19 CAR T cell therapy in diffuse-large B cell lymphoma: Histology and the choice of drug matter

Abstract Introduction Bridging therapy is frequently required during the treatment course of CD19-directed chimeric antigen receptor (CAR) T cell therapy in patients with relapsed or refractory diffuse large B-cell lymphoma (DLBCL). However, its effect on outcome remains unclear and standardized guidelines are lacking. Methods We analyzed a retrospective, multicenter patient cohort treated between January 2019 and May 2025 across 13 German university hospitals. Bridging therapy as treatment between leukapheresis and CAR T cell infusion, but not effects of holding before apheresis were analyzed. Bridging choices were no bridging, intensive chemotherapy (chemo) (e.g. R-DHAP, R-GDP, R-ICE, R-BEAM, or others), gemcitabine/oxaliplatin (R-Gem/Ox), R-Polatuzumab (R-Pola), R-Pola-Bendamustine (Pola-BR), small molecules, radiotherapy (RTX), other historic chemo (ancient), or sequential (=more than one). To better understand bridging effects late after therapy, only 12 months of follow-up were included which excluded bridging with bispecific antibodies due to approval dates. Bridging efficacy just before CAR T was assessed by Lugano criteria as complete response (CR), partial remission (PR), stable disease (SD), or progressive disease (PD). Effects of Ann-Arbor stage, extra nodal disease (END), ECOG, sex, and elevated lactate dehydrogenase (LDH) on outcome were determined in multivariate analyses. Results Bridging therapies were recorded for a total of 471 patients (376 DLBCL not other specified (NOS), 76 transformed follicular lymphoma (trFL), 19 high grade lymphoma (HGL)) who received Axi-cel (64%), Liso-cel (32%), or Tisa-cel (3%). In the entire cohort, histologic subtype, LDH, END, and ECOG significantly separated PFS after CAR T cell infusion. Reflecting the lack of bridging standards, the choice of bridging varied less within a center than between them, even for more recent periods e.g., from 2023. Overall response rate (ORR) to bridging at lymphodepletion (LD) was predictive for PFS as patients in CR showed PFS of 21.8, in PR of 17.9, in SD of 14.9, and in PD of 9.2 mo (p = 0.001). Patients receiving no bridging had a PFS of 15.2 mo. Intriguingly, long-term PFS in trFL was not associated with response to bridging and PFS curves converged over time. We therefore focused our analyses on DLBCL-NOS. Here, response to bridging at LD predicted PFS by Kaplan-Meyer with 20.6 mo in CR, 17.4 mo in PR, 14.4 mo in SD, and 7.8 mo in PD. Interestingly, no bridging patients (22.9%) initially progressed less frequently than most bridged patients but approached the PFS of the SD patient group after 13 months. Except for the ‘ancient’ group, ORR before LD was similar across all bridging groups and could not explain the distinct PFS. Despite similar ORR, the use of R-Gem/Ox (PFS 17.1 mo, 10.6%), Pola-BR (PFS 16.8 mo, 27.1%) and RTX (PFS 18.2 mo, 6.1%) resulted in superior survival in DLBCL NOS patients compared to ancient (PFS 4.5 mo, 5.1%), R-Pola (PFS 10.0m, 4.8%), sequential (PFS 10.5mo, 6.4%), or intensive (PFS 10.9 mo, 12.2%). Importantly, the identified bridging effect on PFS could not be explained by differences in LDH. Furthermore, ‘intensive chemo’ was used more frequently for stage IV disease compared to all other groups (p=0.023) and also in younger patients than the R-Gem/Ox (p=0.001) or the Pola-BR group (p=0.001). The short PFS after R-Pola of 10.0 mo was surprising as it was used in patients with a trend to better ECOG and fewer END sites. The addition of Benda to R-Pola which was typically done after apheresis, improved PFS to 16.8 mo without multivariate bias. Radiotherapy, more common in localized stage and lower END, may explain better outcomes. R-Gem/Ox resulted in a favorable PFS of 17.1 mo, with no bias except for preference in female patients. Conclusions In summary, our highly granular data suggest that the choice of bridging impacts efficacy of CAR T cell therapy beyond mere tumor mass reduction and is independent of LDH. Response to bridging was more relevant in DLBCL-NOS than in trFL. Despite being retrospective and not matched in all characteristics for some bridging groups, certain groups such as RTX, at least in localized stages, or Pola-BR and R-Gem/Ox were favorable. Our data are informative for ongoing bridging strategies and may influence the design of possible future trials.