Outcomes and treatment patterns of patients with primary mediastinal B‐cell lymphoma after CAR‐T cell therapy failure: A DESCAR‐T analysis

Primary mediastinal B-cell lymphoma (PMBL) is a rare subset of aggressive lymphoma with excellent outcome after dose-dense frontline immunochemotherapy but a historically dismal prognosis at relapse.1-4 Leveraging data from the nation-wide prospective DESCAR-T registry, in which all patients who received commercial CAR-T cells in France as a treatment for a hematologic malignancy were included, we recently showed that relapsed or refractory (R/R) PMBL receiving anti-CD19 CAR-T cell infusion exhibited a high complete response (CR) rate and satisfactory survivals.5 This was particularly meaningful for the 62 patients treated with axicabtagene-ciloleucel (axi-cel), for whom 2 years progression-free survival (PFS) and overall survival (OS) rate reached 70.4% and 86.9%, respectively, and compared favorably with more than 1000 other patients with R/R diffuse large B-cell lymphoma (DLBCL) receiving the same treatment. These data, along with other evidence,6-8 strongly support anti-CD19 CAR-T cells as current standard-of-care for R/R PMBL. However, PMBL patients experiencing CAR-T cells therapy failure still represent an unmet clinical need and their outcomes have not been described so far. Albeit a previous work from DESCAR-T registry did describe the very poor OS of patients with DLBCL progressing after CAR-T cell therapy,9 recent changes in the treatment paradigm of R/R PMBL could translate into unexpected outcomes even in very advanced lines and their clinical course need to be specifically studied. Notably, anti-PD1 checkpoint inhibitors (CPI) demonstrated promising efficacy in heavily pre-treated PMBL, and association with brentuximab-vedotin (Bv) may even increase response rate and survivals.10-13 To address this question, we analyzed the clinical course of patients experiencing treatment failure in the CARTHYM study, whose procedures have been previously described.5 Briefly, it leveraged the data of all adults patients with a diagnosis of PMBL confirmed by an expert hematopathologist from the LYMPHOPATH network14 who received an anti-CD19 CAR-T cell infusion in France within the prospective DESCAR-T registry (NCT04328298), between January 2018 and February 2024. All patients provided written informed consent. The study was conducted in accordance with the Declaration of Helsinki and approved by the local ethics committees of each participating center. The CARTHYM study included 82 patients with R/R PMBL treated across 21 centers: 62 received axi-cel, 14 received tisagenlecleucel (tisa-cel), and 6 received lisocabtagene maraleucel (liso-cel). After a median follow-up of 21 months, 27 progression events were observed. These 27 patients, treated across 13 different centers, constitute the study population analyzed in the present report. Characteristics of the 27 patients are described in Table 1. Fourteen patients (52%) had received axi-cel, 11 (40%) tisa-cel, and 2 (7%) liso-cel prior to disease progression. Median age at CAR-T cell infusion were 33 years and sex ratio (M/F) was 1.1. Twelve patients (44%) were treated before 2021 and 15 (56%) between 2021 and 2024. Median time from CAR-T cell infusion to disease progression was 2 months (range 0–5). Progression occurred within 30 (±5) days in 10 patients (37%), between 1 and 3 months in 14 patients (52%), and after 3 months in 3 patients (11%). No relapse occurred after 6 months. At the time of relapse, 16 patients (59%) had advanced-stage disease (Ann Arbor stage III–IV), and 17 patients (74%) elevated LDH levels. All n = 27 After Axi-cel n = 14 After Tisa-cel n = 11 After Liso-cel n = 2 A biopsy was performed at relapse in 12 patients (44%). Pathological findings suggested recurrence of PMBL in 7 out of 12 case, while DLBCL histology was found in 1 cases and grey-zone lymphoma (with features between DLBCL and Hodgkin lymphoma) in 2 cases. Of note, for each of these 3 cases, both the diagnostic and relapse biopsies were reviewed by the same expert hematopathologist. Two cases showed tumoral necrosis. CD19 was positive in 7 out of 8 cases, CD20 was positive in 10 out 11 cases, and PDL1 was positive in 3 out of 3 cases tested. At time of disease progression after CAR-T cells and after a median follow up of 34 months (6−71), one year (1 y) OS was 48.1% for the whole cohort, with a median OS of 11.7 months (Figure 1A). OS was not affected by era of progression (Supporting Information S1: Figure 1). Overall, 22 (82%) patients received a subsequent therapy after CAR-T cells failure while 5 (18%) experienced rapid clinical deterioration and died before any further treatment. Notably, 8 out of 12 patients experiencing very early progression received subsequent treatment, versus 14 out of 15 patients relapsing more than one month after infusion. These treatments were categorized as follow: CPI-based in 11 patients (pembrolizumab or nivolumab), in association with Bv in 4 of them; chemotherapy-based in 1 patient (Rituximab-bendamustine + polatuzumab-vedotin); anti-CD3/CD20 bispecific antibodies in 6 patients (glofitamab, epcoritamab, or plamotamab); and miscellaneous others in 4 patients (Bv-lendalidomide in one, rituximab-lenalidomide in one, or radiotherapy alone in two). Five patients received radiotherapy alone (2 patients) or as consolidative treatment (3 patients). Two patients received consolidative high-dose chemotherapy followed by autologous stem cell transplantation and 1 patient received consolidative allogeneic stem cell transplantation, all 3 while in CR. Best overall response after salvage treatment was complete response (CR) in 9 out of 22 patients (41%) and partial response (PR) in 1 patient (5%) while 12 patients (55%) showed no response, that is, stable or progressive disease. CR was achieved in 7 out of 11 patients treated with CPI-based regimen, 1 out of 6 patients treated with bispecific antibodies and 1 patient treated with Bv + lenalidomide who subsequently underwent ASCT. A swimmer plot of the 22 patients receiving treatment after CAR-T cells failure is shown in Supporting Information S1: Figure 2. Of note, one GZL failed to bispecific antibody but subsequently achieved a sustained remission after a latter line with CPI, and the only one CR achieved with bispecific antibody was in the patient with DLBCL diagnosis at progression. One year PFS and OS of treated patients were 40.9% and 59.1% respectively (Figure 1B,C). The 7 patients who received CPI-based salvage therapy and achieved CR were alive without progression at 2 year. One year PFS of patients treated with CPI or with non-CPI-based regimen were 63.6% and 18.2%, respectively (Figure 1D). Their clinical characteristics are shown in the Supporting Information S1: Table 1. This study is the first to address the specific outcomes of R/R PMBL experiencing failure after CAR-T cell therapy. It exhibits three main findings of clinical significance for daily practice. First, progressions after CAR-T cell infusion, when occurring, are very early events (89% within 3 months and 100% within 6 months, with a median time from infusion of 2 months versus ≈3 months in DLBCL9, 15) of poor prognosis overall. Despite their young age almost 20% of patients died without receiving any subsequent lines of therapy, highlighting the critical need for a very close follow-up after infusion and for a very prompt action at relapse. On another hand, our previous study demonstrated that residual disease at M + 1 evaluation (notably Deauville score 4) may not affect outcome.5 Thus, clinician may face valid issues in rapid decision-making during the critical period of first months following infusion. Second, apparent evolution of histology at relapse may have important therapeutic implications, as 3 out of 12 biopsied patients showed other WHO-defined diagnosis—namely DLBCL and grey-zone lymphoma. One could emphasize that these entities often exhibit close histologic features with PMBL, leading to potential erroneous primary diagnosis, or that stringent diagnosis of PMBL at relapse may be impeded by alterations of surface markers/tumor microenvironment16 following multiple progression. Yet this observation raise serious therapeutical challenges since patients with frank DLBCL histology should reasonably be addressed to bispecific antibodies treatment (their standard-of-care after CAR-T cell failure17-19), while PMBL or even GZL20 would more likely benefit from CPI-based treatment (see below). These data support new histological examination whenever possible, as well as the advice of an expert hematopathologist. Third, a subset of patients fit to receive subsequent therapy after CAR-T cells failure exhibit uncommon good outcomes that seems to abrogate the dismal survival described in LBCL experiencing early progression after infusion.9, 17, 21, 22 These results are notably driven by the sustained response achieved after CPI treatment (long-lasting remission in 7 out of 11 patients). CPI alone or in combination with Bv have already been shown to produce CR rate as high as 40% in the setting of R/R PMBL, associated with a strickling plateau in the survival curves of responder patients,10-13 but data in patients failing to CAR-T cells are lacking. One study showed that CPI following CAR-T cells therapy failure was largely ineffective in LBCL overall, but CR has been achieved in 2 out of 8 PMBL cases.23 Three out of 4 PMBL patients also achieved CR with CPI after CAR-T cell therapy failure in a single-center experience.24 Our real life data suggest that sustained responses may be achieved with CPI in patients with PMBL progressing after CAR-T cell infusion, despite an expected immunosuppression following multiple lines a chemotherapy and lymphodepletion. Therefore, and as emphasized by recent expert recommendations from the LYSA group,25 CPI may represent a valid options for R/R patients failing to both conventional immunochemotherapy and CAR-T cells. However the use of CPI in PMBL remains off-label in most countries including France. In conclusion, this study defines the clinical patterns of PMBL relapse after CAR-T cell therapy and support the use of CPI–based regimens as a preferred therapeutic option in this setting. Given the young age of most patients and the frequent off-label use of CPI across many countries, these findings underscore the need for a structured, multidisciplinary expert platform—including clinical hematologists, nuclear medicine specialists, and hematopathologists—to support timely, individualized treatment decisions in these rare yet potentially curable situations. The authors thank the patients who accepted to be included in the DESCAR-T study, and all the investigators and their staff involved in the data collection and analyses. They also thank Dr Loic Renaud who reviewed this manuscript on behalf of the LYSA scientific committee. The authors thank everyone from the Lymphoma Academic Research Organization (LYSARC) DESCAR-T study group who actively participated in the study. The DESCAR-T registry was partly funded by Gilead, Bristol Myers Squibb, and Novartis; however, they did not participate in the study design, data collection, statistical analysis, or interpretation, and did not provide assistance for manuscript writing or any editorial support. Jean Galtier: Conceptualization; investigation; writing—original draft; methodology; validation; data curation; formal analysis. Pierre Sesques: Writing—review and editing; validation; investigation. Vivien Dupont: Software; formal analysis. Emmanuel Bachy: Validation; writing—review and editing. Roberta Di Blasi: Validation; writing—review and editing. Catherine Thieblemont: Validation; writing—review and editing. Gabriel Brisou: Validation; writing—review and editing. Thomas Gastinne: Validation; writing—review and editing. Guillaume Cartron: Validation; writing—review and editing. François-Xavier Gros: Validation; writing—review and editing. Franck Morschhauser: Validation; writing—review and editing. Axel André: Validation; writing—review and editing. Jacques-Olivier Bay: Validation; writing—review and editing. Magalie Joris: Validation; writing—review and editing. Aline Tanguy-Schmidt: Validation; writing—review and editing. Audrey Demailly: Resources; project administration. Steven Le Gouill: Validation; writing—review and editing. Roch Houot: Conceptualization; validation; writing—review and editing. Krimo Bouabdallah: Conceptualization; validation; writing—review and editing. Vincent Camus: Conceptualization; validation; writing—review and editing; supervision; investigation; methodology. The authors declare no conflicts of interest. This research received no funding. Data are subject to controlled access by DESCAR-T owing to privacy and legal requirements and proprietary reasons. Anonymized IPD requests will be promptly reviewed by the corresponding author (J.G.) and the scientific committee of the DESCAR-T registry. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.