Day +3/Day 0 Ferritin Ratio: A Simple Point-of-Care Index Predictive of Severe Immune Effector Cell-Associated Neurotoxicity Syndrome after CD19 CAR T-Cell Therapy

Introduction: Chimeric antigen receptor (CAR) T-cell therapy is transforming the management of patients with relapsed/refractory (R/R) B- and plasma-cell malignancies. However, a major limitation of CAR T-cell therapy is the development of severe, sometimes life-threatening, toxicities such as immune effector cell-associated neurotoxicity (ICANS) and cytokine release syndrome (CRS). While the management of CRS has dramatically improved with the early use of tocilizumab and corticosteroids, how to best predict and treat ICANS remains poorly understood, with up to 30% of patients developing severe ICANS after CD19 CAR T-cell therapy. Since we had previously observed features overlapping with hemophagocytosis lymphohistiocytosis in patients with severe CRS and ICANS, we hypothesized that the early kinetics of ferritin and soluble IL-2 receptor alpha [sIL-2Rα] post infusion could predict the development of severe ICANS. Methods: We retrospectively analyzed data from 195 patients with R/R B-cell malignancies previously treated on a phase I/II clinical trial (NCT01865617) of an investigational defined 1:1 CD8+:CD4+ composition CD19 CAR-T product (JCAR014; training set). CRS and ICANS severity were graded using Lee 2014 and CTCAE 4.03 criteria, respectively. Logistic regression was used to predict the occurrence of severe ICANS. Ferritin was measured using an enzyme immunoassay. sIL-2Rα was measured using Luminex®. Goodness of fit was assessed using the Akaike information criterion (AIC). Independent test set included 203 patients treated at our center with the standard-of-care CD19 CAR T-cell products axicabtagene or lisocabatagene maraleucel. Discrimination and calibration were assessed in both the training and test set. Results: In the training set (n=195), median age was 55 (range: 20-76). Disease type was as follows: acute lymphoblastic leukemia, n=64 (32.8%); B-cell non-Hodgkin lymphoma, n=82 (42.1%); chronic lymphocytic leukemia, n= 47 (24.1%). We observed any grade and grade ≥3 CRS in 134 (68.7%) and 27 (13.8%), respectively, and any grade and grade ≥3 ICANS in 72 (38.8%) and 37 (19.0%), respectively. Longitudinal analyses of serum ferritin (Figure 1) and sIL-2Rα values using locally estimated scatterplot smoothing suggested that the rate of increase of both analytes was markedly higher in patients with grade ≥3 ICANS compared to grade 0-2 ICANS. We applied univariate logistic regression to estimate the association between serum ferritin levels at early timepoints, select ratios of ferritin and sIL-2Rα values, and the occurrence of severe ICANS. The strongest association was observed between the ferritin day +3/day 0 ratio and severe ICANS (OR per log10 increase, 175; 95% CI, 19.1-2668; p < 0.001). We could further improve our model by modeling the ferritin day +3/day 0 ratio with a restricted cubic spline to allow for non-linear effects (unsplined model, AIC = 129, C-index, 0.74; splined model, AIC = 127, C-index, 0.78). Ferritin day +3/day 0 ratios of 3, 5, and 10 were associated with an 81%, 95%, and 99% probability of grade ≥3 ICANS, respectively (Figure 2). The splined model was associated with near-perfect calibration corrected for optimism using bootstrapping. We next evaluated the predictive ability of the ferritin day +3/day 0 ratio to predict severe ICANS in our independent test set. Predictions in the test set based on our splined model retained good discriminative ability with a C-index of 0.64, although with suboptimal calibration. In a refitted logistic regression model in the test set, the ferritin day +3/day 0 ratio remained strongly associated with the development of severe ICANS (OR = 18.8 per log10 increase, 95%CI, 2.48-142.01, p = 0.004). Conclusion: We validated the ability of a simple point-of-care index (day +3/day 0 ferritin ratio) to predict the development of severe ICANS after CD19 CAR T-cell therapy in our training and test set. Since serum ferritin levels can be quickly measured by most clinical laboratories, the day +3/day 0 ferritin ratio could become a practical and useful tool for practitioners taking care of CAR T-cell patients. The early identification of patients at high risk of developing severe ICANS days prior to symptoms onset could enable the evaluation of targeted prophylactic interventions in future clinical trials. We plan to refine our modeling approach to further improve the calibration in test set.