Comparison of axicabtagene ciloleucel to standard regimens as second line treatment for large B cell lymphoma in real life: A lysa study from descar-T and realysa registries

Abstract

Abstract IntroductionThe treatment landscape of early relapsed/refractory (R/R) large B-cell lymphoma (LBCL) has been transformed by anti-CD19 CAR T-cells. In ZUMA-7, transplant-intended early R/R LBCL patients treated with axicabtagene ciloleucel (axi-cel) had longer event-free survival (EFS) than those receiving standard salvage chemoimmunotherapy (SOC). At 47.2 months' median follow-up, median overall survival (OS) was not reached for axi-cel and was 31.1 months for SOC; estimated 4-year OS was 54.6% and 46.0%, respectively (Locke NEJM 2022, Westin NEJM 2023). Since approval, axi-cel has become the predominant CAR T therapy in France for 2L treatment since July 2022. However, access remains uneven, as CAR T-cell therapy is not universally available. In this context, we performed a retrospective multicenter study to assess real-world outcomes of early R/R LBCL patients treated with 2L axi-cel versus SOC. MethodologyTo quantify the effect of intention to receive axi-cel versus SOC as 2L therapy in early R/R LBCL, we analyzed individual-level data from two French studies: the REALYSA cohort (SOC) and DESCAR-T registry (axi-cel). Adults (≥18 years) with LBCL, high-grade B-cell lymphoma (HGBL), or primary mediastinal B-cell lymphoma (PMBL) refractory to first-line immunochemotherapy or relapsed within 12 months were included, regardless of transplant eligibility. To minimize selection bias, we included REALYSA patients treated in 2L before axi-cel's 2L approval (<July 2022), and axi-cel cases treated in 2L from July 2022 to December 2023 with at least 12 months' follow-up. Time-to-event outcomes were measured from the date of strategy decision (axi-cel) or 2L start (SOC). Groups were balanced with propensity scores (PS) and stabilized inverse probability weighting (sIPTW) using age (spline), sex, histology, stage, ECOG performance status (PS), LDH, and time from diagnosis to 2L. Outcomes were EFS (to progression, new therapy, or death) and OS. The weighted Kaplan–Meier estimator was used. ResultsWe identified 659 eligible patients: 465 received axi-cel, 194 SOC. After excluding those with missing covariate data, the complete case analysis included 356 axi-cel and 132 SOC patients. At baseline: most had advanced stage (82% axi-cel, 80% SOC), and primary refractory disease (77.5% axi-cel, 77.3% SOC). Mean time from diagnosis to 2L was similar (0.63 years axi-cel, 0.64 years SOC). SOC patients were older (mean 65.9 vs 58.6 years). HGBL (18.2% vs 5.1%) and PMBL (6.1% vs 3.7%) were more frequent in SOC, and ECOG PS ≥2 was higher in SOC (30% vs 12%). SOC regimens were mainly platinum-based (R-DHAC/Ox 38%, R-GemOx 21%). In the SOC arm, 35.6% were <65 years (vs 57.6% axi-cel); 10.8% underwent autologous transplant (24.6% among those ≤65 years). Median follow-up was 1.2 years (axi-cel) versus 2.8 years (SOC). After PS-weighting, baseline variables were well-balanced (standardized mean differences <0.1). EFS at 1 year favored axi-cel: 46% vs 16% (weighted log-rank p<0.001). OS showed a trend favoring axi-cel but was not statistically significant (1-year OS 64% vs 57%, p=0.46). In SOC, 62.4% (n=121/194) received third-line therapy, mostly CAR T-cells (n=49/121, 40.5%) or salvage chemotherapy (n=43/121, 35.5%); lenalidomide-based regimens were given in 18.2% (n=22/121), and only 2 patients received bispecifics. ConclusionIn this large French cohort, including both transplant-eligible and ineligible early R/R LBCL patients, a clear EFS benefit was observed with axi-cel, consistent with ZUMA-7 results. However, this did not translate into a significant OS benefit, possibly due to short follow-up or third-line CAR T use in the SOC arm. Indeed, CAR T was the most frequent third-line option after SOC and may have impacted OS. Further analysis is needed to clarify the optimal role of CAR T-cell therapy in R/R LBCL, including medico-economic evaluation. Ongoing work will address missing data with multiple imputations for improved estimates.