Identification of Predictive Markers of Severe and Prolonged Neutropenia after CD19-Specific CAR T-Cell Treatment in Patients with Relapsed/Refractory B-Cell Malignancies

Introduction: Genetically engineered CD19-specific CAR T-lymphocytes have emerged as a powerful new class of immunotherapeutic agents in relapsed/refractory B-cell malignancies. However, their utility is hampered by unique toxicities including prolonged cytopenias that can predispose for severe infectious complications. Here, we evaluate predictive markers of hematotoxicity in patients treated in the real-world setting.

Methods: To define the incidence and nature of CAR T-cell associated hematotoxicity, we performed a retrospective chart review across three German academic hospitals (LMU Munich, Tübingen, Hamburg-Eppendorf). Between January 2019 and June 2020, a total of 69 patients (58 r/r B-NHL, 11 BCP-ALL) were studied following treatment with the commercial CD19-specific CAR T-cell products Axi-cel (n=43) or Tisa-cel (n=26). The relative impact of baseline (=prior to lymphodepletion) demographic, laboratory and clinical characteristics on hematotoxicity was analyzed by uni- & multivariate analysis using the following primary endpoint: duration of clinically significant neutropenia – defined as an ANC <500/µl – between days 0-60.

Results: In this heavily pre-treated cohort, the incidence of severe neutropenia (ANC <500/µl) was 97% with a median duration of 12 days (95% CI: 10-15). Neutropenia followed a bimodal temporal curve with intermittent hematopoietic recovery upon G-CSF support. Profound, protracted (ANC <100/µl for ≥7 days) and prolonged (ANC <1000/µl after day 21) neutropenia was common, occurring in 36% and 78% of patients respectively. The incidence of severe anemia (Hb <7 g/dl or requiring pRBC transfusion) and severe thrombocytopenia (PLT count <50 G/l) was 68% and 71% respectively. Of note, a total of 12 patients (17 %) observed severe, G-CSF-refractory bone marrow aplasia with an ANC < 500/µl greater than 30 days.

The following baseline clinical markers were significantly correlated with the duration of neutropenia by univariate or logistic regression analysis (Figure 1): Platelet Count (r = -0.45, P=0.0001), Hemoglobin (r = - 0.33, P=0.006), ANC (r = - 0.28, P=0.02), C-reactive Protein (r = +0.35, P=0.004), Ferritin (r = +0.52, P= <0.0001), and BM Infiltration for lymphoma (LR=6.7, P=0.0096). Interestingly, the following baseline factors were not significantly correlated: LDH, age, prior lines of cytotoxic therapy, and eGFR. On stepwise multivariate analysis using a binary logistic regression for the outcome severe neutropenia ≥ 14 days, baseline ferritin, hemoglobin and platelet count retained significance.

In terms of dynamic clinical variables, both peak IL-6 and peak CRP were not associated with increased hematotoxicity, though peak IL-6 exhibited a trend towards significance (r = +0.22, P= 0.077). While peak ferritin was associated with a longer duration of neutropenia (r = +0.37, P=0.025), the peak value was less predictive than the baseline value. Moreover, we observed no nested dependency between hematotoxicity and CRS/ICANS. CRS grade in particular was not associated with prolonged neutropenia in our study cohort. Furthermore, the application of G-CSF was not associated with an increased rate or severity of neurotoxicity. When comparing the duration of neutropenia by CAR product (Axi-cel vs. Tisa-cel) and disease entity (DLBCL vs. BCP-ALL), no statistically significant difference was observed.

Conclusions: Profound and prolonged neutropenia was common in our study cohort. This included cases of severe, G-CSF-refractory bone marrow aplasia. The identified predictive markers point toward a multifactorial model of pathogenesis for CAR T-cell associated hematotoxicity. They include markers of impaired hematopoietic reserve (e.g. low baseline PLT count, Hemoglobin, ANC), inflammatory state (e.g. hyperferritinemia, elevated CRP), and tumor microenvironment (e.g. BM infiltration). These findings carry important implications for both risk-stratification and the development of predictive clinical tools. The identified factors are currently being validated in an independent patient cohort and prospective validation is ongoing.