Comparative CRS risk across CAR-T constructs in DLBCL in real-world cohorts

Abstract Background Cytokine release syndrome (CRS) is a common and potentially life-threatening toxicity associated with chimeric antigen receptor T-cell (CAR-T) therapy. Although clinical trials have demonstrated varying CRS profiles across different CAR-T constructs, real-world comparative data remains limited.Our study aims to evaluate CRS risk across commercial CAR-T therapies in a real-world DLBCL cohort, with particular focus on the timing of inflammatory biomarkers and the use of tocilizumab as a proxy for CRS management. Methods: We conducted a retrospective analysis using the TriNetX Global Collaborative Network. Adult patients (≥18 years) with diffuse large B-cell lymphoma (DLBCL) who received FDA-approved CAR-T therapies axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), and lisocabtagene maraleucel (liso-cel) were included. For each therapy, two cohorts were created: patients with elevated inflammatory markers prior to infusion (within 1 month) and post-infusion (within 14 days). Elevated markers were defined as ferritin ≥100 ng/mL, LDH ≥250 U/L, and CRP ≥30 mg/L. These biomarkers were also evaluated as secondary outcomes to compare the degree of systemic inflammation across CAR-T constructs.Propensity score matching (1:1) was performed to balance age, race, chemotherapy lines, glucocorticoid use, diabetes mellitus type II,congestive heart failure and chronic kidney disease (CKD). The primary outcome was the diagnosis of cytokine release syndrome (CRS) within 90 days post-CAR-T. Secondary outcomes included post-infusion ferritin and CRP levels, and use of tocilizumab as a surrogate for CRS severity. We conducted both within-product comparisons (pre- vs post-infusion inflammatory states) and cross-product comparisons between axi-cel, tisa-cel, and liso-cel. Survival and risk analyses were performed using Kaplan-Meier curves and measures of association. Results: Among DLBCL patients with elevated post-infusion inflammatory markers, CRS incidence varied across CAR-T constructs but did not differ significantly in matched comparisons. In within-product analyses, CRS occurred in 55.6% of axi-cel recipients pre-infusion versus 47.6% post-infusion (risk difference +7.9%, p=0.621). For tisa-cel, CRS occurred in 100% of patients in both cohorts. In liso-cel recipients, CRS occurred in 58.8% pre-infusion versus 71.4% post-infusion (risk difference –12.6%, p=0.465). In post-infusion biomarker-positive cohorts with adjusted matching, CRS occurred in 83.3% of axi-cel vs 66.7% of tisa-cel patients (risk difference +16.7%, p=0.326), and in 71.4% of axi-cel vs 76.9% of liso-cel recipients (risk difference –5.5%, p=0.745). None of these differences reached statistical significance. Kaplan-Meier analyses revealed no difference in CRS-free survival between groups. Tocilizumab was used more frequently in axi-cel patients compared to tisa-cel (52.6% vs 41.7%, p=0.474) and liso-cel (58.8% vs 38.5%, p=0.191). Among patients with post-infusion biomarker positivity, ferritin levels were numerically higher in liso-cel compared to axi-cel (3311 vs 2417 ng/mL, p=0.726), and significantly higher in axi-cel than tisa-cel (8806 vs 1551 ng/mL, p=0.216). CRP levels were significantly higher in tisa-cel vs axi-cel (34.4 vs 9.5 mg/L, p=0.011), and numerically higher in liso-cel vs axi-cel (9.7 vs 4.9 mg/L, p=0.132). These biomarker trends suggest construct-specific inflammatory responses despite similar CRS incidence. Conclusion: In this real-world, propensity-matched analysis of DLBCL patients with elevated post-infusion inflammatory markers, we observed high rates of cytokine release syndrome (CRS) across all CAR-T constructs. While CRS incidence was numerically higher with axi-cel compared to tisa-cel and slightly lower compared to liso-cel, these differences were not statistically significant. Importantly, we found that inflammatory biomarker levels particularly CRP and ferritin varied significantly across products, despite comparable CRS event rates. Collectively, these findings suggest that while CRS rates may appear similar across CAR-T products in high-risk inflammatory states, the underlying inflammatory burden—as reflected by ferritin and CRP—may differ by construct. Future studies should consider biomarker trajectory and comorbidity burden when developing risk-adapted toxicity management strategies. These insights support more nuanced post-infusion monitoring and product-specific counseling in clinical practice.