Site-Specific Analysis of Extranodal Involvement Reveals Distinct Efficacy Outcomes with CAR T-Cell Therapy in Large B-Cell Lymphoma Patients

Introduction: While chimeric antigen receptor (CAR) T-cell therapy represents the standard-of-care for patients with relapsed/refractory (R/R) large B-cell lymphoma (LBCL), only 40% achieve a durable remission. Initial reports point towards worse efficacy outcomes in patients with extranodal involvement, but large cohorts with granular information on number and specific site data are lacking. Here, we comprehensively mapped peri-CART extranodal involvement and analyzed its prognostic impact in a broad, real-world patient population. Methods: We conducted a multicenter, international study including patients with R/R LBCL treated at 8 centers with commercially-available CAR T-cell products until May 2024. Lesion-level description of extranodal involvement was performed in pre-CART scans for all patients, and at relapse for those experiencing progressive disease (PD) after CAR-T. Safety and efficacy outcomes according to the presence of extranodal disease (ED) at time of CAR-T were analyzed. In the second part of the study, focused on the ED cohort, we carried out a site-specific analysis to determine the impact of the number and ED location on efficacy after infusion. Finally, we analyzed patterns of relapse in patients who experienced PD after CAR-T, aiming to identify CART-refractory extranodal sites. Results: A comprehensive review of nodal and extranodal involvement was performed in 800 scans of 516 patients with R/R LBCL treated with axicabtagene ciloleucel (65%) or tisagenlecleucel (35%). Of these, 339 (66%) patients had ED at time of CAR-T, with (N=273, 81%) or without (N=66, 19%) concomitant nodal involvement (ED cohort). Regarding baseline characteristics, the ED cohort included more patients with ECOG >1 (14% vs 3%, p<0.001), bulky disease (31% vs 18%, p=0.003), high LDH (65% vs 51%, p=0.011) and a high CAR-HEMATOTOX score (58% vs 38%, p<0.001) in comparison to the only-nodal disease (ND) group. The remaining variables were balanced, including CAR-T construct use. First, we analyzed the impact of ED on efficacy and safety after CAR T-cells. Patients in the ED cohort presented higher rates of neurotoxicity any grade and grade >2 than ND patients (45% vs 35% [p=0.030] and 19% vs 11% [p=0.016]), without differences in cytokine release syndrome. Concerning efficacy, the ED cohort had a lower overall (68% vs 82%, p<0.001) and complete response rate (CRR, 47% vs 62%, p<0.001) in comparison to ND patients. With a median follow-up from infusion of 24 months, median progression-free survival (PFS) and overall survival (OS) were 4.6 vs 9.2 months (p<0.001) and 12 vs 32 months (p<0.001) for the ED and ND cohorts, respectively. In the multivariable analysis, presence of ED at time of CAR-T retained a significant impact on PFS (p<0.01) and OS (p=0.03). In the second part of the study, we focused on the ED cohort (N=339). Median number of sites was 1 (IQR 1-2) and the most frequent locations were bone (110, 32%), soft tissue (107, 32%), lung (75, 22%), liver (47, 14%), pleura (46, 14%) and gastro-intestinal (39, 12%). Patients with >1 ED site presented a shorter PFS (4.4 vs 5.3 [p=0.041]) and OS (8.4 vs 17.0 [p=0.004]) in comparison to patients with 1 ED site. Interestingly, patients with only ED had a significantly shorter PFS (median 4.5 vs 7.4 [p=0.02]) and OS (median 10 vs 24 months [p=0.004]) in comparison to patients harboring concomitant nodal involvement. Considering specific ED location, CRR was heterogeneous: 50% in soft tissue, 48% in pleura, 44% in bone, 39% in lung, 39% in liver and 35% in gastro-intestinal involvement. Interestingly, there was a low CRR in infrequent locations such as pancreas (8/23 [35%]), adrenal glands (6/22 [27%]), pericardium (0/6) and uterus (1/5 [20%]), while patients with CNS involvement achieved a 67% CRR (6/9). Of the ED cohort, 190 (56%) patients experienced PD during follow-up, most of them with ED involvement (91%) in the same location present prior to CAR-T. Certain ED sites showed a numerically higher risk of relapse, such as bone (14/51, 27%), pleura (6/22, 27%), liver (6/19, 32%), and CNS (4/6, 67%). Conclusions: Leveraging longitudinal lesion-level data of extranodal involvement from a large international cohort, we uncovered site-specific patterns of response and relapse to CAR T-cells. Both a progressively higher number of extranodal sites and certain disease locations, such as liver and adrenal glands, were associated to reduced efficacy after infusion.