Which inflammation matters? distinct risk signatures before CAR-T cell therapy and the role of inflamix

Abstract Car T therapy improved the prognosis of relapse/refractory lymphoma patients, but managing toxicities and progressions remains challenging. Scores based on inflammation and peripheral blood counts, such as modified EASIX (Endothelial Activation and Stress Index) and CAR-HEMATOTOX, were designed to predict toxicities as cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) . In order to identify patients at high risk of CAR T treatment failure Raj S.S. et al. developed InflaMix (Inflammation Mixure Model), a quantitative model integrating 14 pre-CAR-T infusion laboratory measures that revealed patients at high risk of CAR-T failure in terms of PFS and OS. In this study, we aim to validate the InflaMix score in patients undergoing CAR-T therapy . Materials and Methods We analysed adult LBCL patients who received CAR T-cell therapy at our institution between September 2019 and March 2025. We set the data cut-off at June 2025 to ensure full documentation of early toxicities (up to day +30) and assessment of disease status by PET-CT within three months for all enrolled patients. Using pre-infusion values, we classified patients as “inflammatory” (score ≥1, n = 21) or “non-inflammatory” (n = 71) based on predefined InflaMix criteria. We tested the correlation of the InflaMix score with progression-free survival (PFS) and overall survival (OS). Then, we then correlated the InflaMix score with the onset of early toxicities. Next, we evaluated potential differences in the use of tocilizumab, steroid therapy, and intensive care unit (ICU) hospitalization between the two groups. Finally, we determined if there were differences in the development of coagulopathy and neutropenia between the two groups. Additionally, we calculated the mEASIX and CAR-HEMATOTOX scores for all patients using pre-lymphodepletion values. All statistical analyses were conducted using NCSS 2020 software. Results This analysis included 92 patients with LBCL who all received anti-CD19 CAR T-cell therapy. The median age was 59 years (range 56-62) in the non-inflammatory group and 54 years (range 47-60) in the inflammatory group. Most patients in both groups were male (56% and 52% p=ns). The median number of prior therapy lines was two in both groups (range 2–3). Among the “inflammatory” patients, 4 (19%) had controlled disease (complete or partial response), and 7 (8%) had lactate dehydrogenase (LDH) within limits prior to lymphodepletion therapy. Among the “non-inflammatory” group, 28 patients (40%) had controlled disease (complete or partial response), and 20 patients (28%) had normal LDH levels prior to lymphodepletion therapy. Overall, 65 patients (71%) received axicabtagene ciloleucel (axicel), and 27 patients (29%) received tisagenlecleucel (tisacel). Calculating the mEASiX revealed that it was significantly higher in the inflammatory group (median 5.82, range 3.98–6.97) than in the non-inflammatory group (median 2.88, range 1.04–3.82) (p < 0.001, Mann–Whitney U test). There was no statistically significant difference when comparing the two groups using the CAR-HEMATOTOX score (p = 0.555). Next, we evaluated whether there was a statistically significant difference in progression-free survival (PFS) and overall survival (OS) between the inflammatory and non-inflammatory groups. We found statistically significant differences in PFS (median 196 vs 76 days , p = 0.002) and OS (median 275 vs 198 = 0.039), with hazard ratios of 2.74 (1.46–5.16) and 2.15 (1.03–4.46), respectively. There was no statistically significant difference between the two groups in terms of grade 2-4CRS (69% vs 67%, p = 0.838), or any ICANS (77% vs 76%, p = 0.786). Additionally, we found no statistically significant difference in the use of tocilizumab (p= 0.5) to treat CRS or steroid (p= 1) therapy to treat ICANS. Patients in both groups showed no statistically significant difference in the development of coagulopathy or neutropenia (p= 0.85). Finally, we found no differences in ICU admission (p= 0.34). Conclusions We validate the InflaMix score in a European real-world setting as an independent prognostic tool that identifies high-risk patients prior to CAR-T cell infusion, not predicting early toxicities. Its divergence from mEASIX and CAR-HEMATOTOX suggests that it reflects a distinct inflammatory profile. This underscores the importance of incorporating multiple biomarkers to enhance risk stratification.