Real-time interleukin-6 (IL-6) kinetics predict cytokine release syndrome (CRS) in patients receiving chimeric antigen receptor (CAR) T-cell therapy for Relapsed/Refractory B-cell malignancies

Abstract

Abstract Background Cytokine release syndrome (CRS) and Immune Effector Cell Associated Neurotoxicity Syndrome (ICANS) are common, potentially life-threatening toxicities following CAR T-cell therapy. Elevated IL-6 is mechanistically linked to CRS, but most studies have retrospectively analyzed static or peak cytokine levels. We hypothesized that prospective, real-time, IL-6 measurement is feasible in patients (pts) receiving CAR T cells, and that IL-6 kinetics in the early post-infusion period could predict CRS and ICANS. Methods We prospectively enrolled pts with relapsed/refractory B-cell non-Hodgkin lymphoma (NHL), or multiple myeloma (MM) treated with CAR T-cells. Serum samples were drawn at pre-defined time points [apheresis, day -5 of lymphodepletion (LD), day of CAR T infusion (day 0), 12 hours post infusion (day 0.5), and days 1, 2, 4 & 7] to run IL-6 assay (Beckman Coulter) in real-time. IL-6 log-slope was calculated using log-transformed values from day 0 to the last sample preceding onset of CRS; excluding samples drawn after CRS onset or anti IL-6/IL-6R therapy. CRS and ICANS were graded per ASTCT criteria. Associations between IL-6 values and toxicity were analyzed using nonparametric tests, univariable logistic regression, and receiver operator curve (ROC) analysis. Optimal log-slope cutoff was determined using Youden's index. Results 42 pts were enrolled, 18 (43%) had MM, 24 (57%) had NHL. Median age was 67 years (range 39–84), 88% pts were White, and 19% pts had ECOG ≥2. Sixteen pts (36%) had prior autologous transplant, and one had an allogeneic transplant. CAR T-cell products included ciltacabtagene ciloleucel (43%), lisocabtagene maraleucel (24%), axicabtagene ciloleucel (21%), and brexucabtagene autoleucel (12%). Thirty (71%) pts received lymphodepletion with fludarabine/cyclophosphamide, while 12 (29%) received bendamustine. CRS and ICANS occurred in 30 (71%) and 13 (31%) pts, respectively. Among pts experiencing CRS and ICANS, grade (G) 2 and higher cases were reported in 16 (53%) and 8 (62%) pts, respectively. Median time to CRS and ICANS onset was 5 (range: 1-9) and 6 (range: 3-17) days, respectively. Elevated C-reactive protein (CRP) levels were significantly associated with CRS from day 3 onwards, peaking on day 5 in pts with and without CRS (median 2.0 vs 0.3 mg/dL, p=0.002). Ferritin was elevated in pts who developed CRS but did not reach statistical significance. EASIX scores showed no significant difference by CRS or ICANS status. Median IL-6 turnaround time across all timepoints was 1.7 days (IQR: 0.6–3.5). A higher day 1 IL-6 level was noted in pts. with ≥G2 CRS (median 20 vs. 8 pg/mL; p=0.013), and peak IL-6 levels were higher in pts with any CRS vs. no CRS (median 33 vs. 12 pg/mL; p=0.002) and CRS ≥G2 vs. G0-1 (p=0.005). IL-6 log-slope was significantly higher in pts who developed CRS [median 0.77 vs. 0.30 (range: 0.10-0.41)], p <0.001. When stratified by CRS severity, a higher IL-6 log-slope was seen in pts with ≥ G2 CRS (median 0.92 vs. 0.47 in G0-1 CRS, p=0.009). The IL-6 absolute value on day 1 was higher in pts who eventually developed ICANS [median 19 vs. 8 pg/mL; p=0.007], and so was the IL-6 peak level (median 41 vs 19 pg/mL, p=0.009). IL-6 log-slope was also significantly higher in pts who experienced ICANS vs no. ICANS (median 0.98 vs 0.52; p=0.05) and ≥G2 vs. G0-1 ICANS (median 1.12 vs 0.52; p=0.037). ROC analysis demonstrated that IL-6 log-slope predicts CRS with an area under the curve (AUC) of 0.88. An optimal log-slope cutoff of 0.64 -corresponding to a daily IL-6 increase of ≈90% per day-yielded 89% sensitivity and 92% specificity. For ICANS, ROC analysis demonstrated that IL-6 log-slope predicts ICANS with an AUC of 0.72. A cut-off of 0.25 -corresponding to an IL-6 increase of ≈29% per day- yielded sensitivity and specificity of approximately 67% for predicting ICANS. Conclusions In this prospective pilot study, we demonstrate the feasibility of real-time IL-6 monitoring for patients treated with CAR T cells. We also show that early IL-6 dynamics can predict CRS with high accuracy. These results can be integrated into clinical workflows used to identify patients at high risk of CRS and can enable future studies investigating risk-adapted interventions such as preemptive tocilizumab, hospitalization decisions, and toxicity mitigation strategies. Further validation in larger cohorts is warranted to confirm these findings and inform future practice algorithms.