Large B-cell lymphoma following CD19-directed CART-cell therapy failure: Impact of CAR construct on disease features and outcomes of subsequent therapies

Abstract Introduction: All FDA-approved chimeric antigen receptor (CAR) T-cell therapies targeting CD19 (CART19) for relapsed/refractory (r/r) large B-cell lymphoma (LBCL) utilize the FMC63-derived single-chain variable fragment. However, they differ in co-stimulatory domain (CD28 in axi-cel versus 4-1BB in tisa-cel and liso-cel) which influences T-cell expansion and persistence. Whether these design differences affect the biology of the relapse following CART19 or the efficacy of subsequent therapies remains unclear.Methods: We conducted a retrospective, single-center study of LBCL patients (pts) who were r/r to commercial CART19 products. Data cutoff was June 2025. Clinical data, relapse tumor characteristics, and outcomes of subsequent therapies were collected from pt records. CD19 status was assessed by flow cytometry and/or immunohistochemistry; genomic profiling was retrieved by institutional assay (PENNSEQ hematological panel). Progression-Free Survival-2 (PFS2) was defined from the first post-CART progression to the next event (progression, new therapy, death, or last follow-up).Results: Of 327 LBCL pts treated with CART19 (01/2018–12/2024), 77% (N=251) received a 4-1BB-based product (tisa-cel N=181; liso-cel N=70) and 23% (N=76) the CD28-based axi-cel. CART19 was given in 2nd line in 19.5% of pts (N=64; 29 liso-cel, 35 axi-cel). In the full cohort, median age was 63 years (21-86 yrs); 64% of pts were male, 8% had prior CNS involvement, and 95% had ECOG 2 at CART. The median number of prior therapies was 3 (1-12),14% of pts had prior autologous transplant, 83% received bridging, and 34% had elevated LDH. Baseline characteristics were similar between CART costimulation groups, though pts receiving CD28-based CART were younger (56 vs. 66 yrs; p0.01), more often received fludarabine/cyclophosphamide lymphodepletion, and treated more frequently in 2nd line. With a median follow-up of 30.9 months (95% CI:24.9–36.02), 60.5% of pts (N=198) experienced disease progression, with no significant difference by CART costimulation (p = 0.34). Among those who progressed, 7% (N=13) had isolated CNS relapse, 45% (N=90; 70% of whom were stage IV at infusion) had extranodal progression, and 48% (N=95) had nodal relapse. Of all relapses, 47% (N=94) occurred during the first 3 months, 37% (N=73) during months 3-12, and 16% (N=31) after 12 months with no difference by construct. Among pts with available relapse biopsies (N=97), 68% (N=66) were CD19-bright, 18% (N=17) CD19-dim, and 14% (N=14) CD19-negative, with numerically higher CD19-negative cases in CD28-based CART (21% vs. 12%, p=0.3). Genomic profiling at relapse was available for 57 pts, revealing frequent mutations in KMT2D (34%), TP53 (30%), and CREBBP (25%). CIITA mutations that may lead to overexpression of PD-L1/2 were observed in 9% and were more common after CD28-CART (21% vs. 3%, p=0.04). Interestingly, LZRT1 and PIK3C2B mutations were detected at low frequencies (4/43) exclusively in pts relapsing after CART. The median PFS2 following CART progression was 3.88 months (95% CI: 3.26–5.83). Pts with extranodal relapse (3.4 vs. 5.6 months), elevated LDH (3.0 vs. 6.1 months), or high ferritin (1.9 vs. 6.4 months) at relapse had shorter PFS2 (p 0.01). In multivariate analysis, elevated LDH and ferritin remained independently associated with shorter PFS2. The frequency of these adverse features did not differ by prior CART subtype. PFS2 varied significantly by type of subsequent therapy (polatuzumab-based regimens, bispecific antibodies, lenalidomide-based regimens, radiotherapy /local therapy, targeted agents (Di Blasi et al. Blood 2022), and other treatments; p=0.012). Polatuzumab-based therapies achieved the longest median PFS2 at 7.6 months (bispecific 3.9, lenalidomide 2.7, local 3.1, targeted 4.2, other 2.3 months). However, within these treatment groups, outcomes were similar regardless of the prior CART constructs.Conclusions: In this study of LBCL pts relapsing after CART19, PFS2 was overall short for both 4-1BB and CD28 CART19. CD19-negative relapses and CIITA mutations were more frequent after axi-cel. Polatuzumab-based regimens achieved the longest PFS2. Ongoing analyses are assessing CART-cell persistence and CD19 mutational changes pre-/post-treatment to identify selective pressures. Better insight into relapse biology may inform personalized strategies post-CART in this high-risk population.