CAR-HEMATOTOX: a model for CAR T-cell-related hematologic toxicity in relapsed/refractory large B-cell lymphoma

Abstract

Hematotoxicity represents a frequent chimeric antigen receptor (CAR) T-cell-related adverse event and remains poorly understood. In this multicenter analysis, we studied patterns of hematopoietic reconstitution and evaluated potential predictive markers in 258 patients receiving axicabtagene ciloleucel (axi-cel) or tisagenlecleucel (tisa-cel) for relapsed/refractory large B-cell lymphoma. We observed profound (absolute neutrophil count [ANC] <100 cells per µL) neutropenia in 72% of patients and prolonged (21 days or longer) neutropenia in 64% of patients. The median duration of severe neutropenia (ANC < 500 cells per µL) was 9 days. We aimed to identify predictive biomarkers of hematotoxicity using the duration of severe neutropenia until day +60 as the primary end point. In the training cohort (n = 58), we observed a significant correlation with baseline thrombocytopenia (r = -0.43; P = .001) and hyperferritinemia (r = 0.54; P < .0001) on univariate and multivariate analysis. Incidence and severity of cytokine-release syndrome, immune effector cell-associated neurotoxicity syndrome, and peak cytokine levels were not associated with the primary end point. We created the CAR-HEMATOTOX model, which included markers associated with hematopoietic reserve (eg, platelet count, hemoglobin, and ANC) and baseline inflammation (eg, C-reactive protein and ferritin). This model was validated in independent cohorts, one from Europe (n = 91) and one from the United States (n = 109) and discriminated patients with severe neutropenia ≥14 days to <14 days (pooled validation: area under the curve, 0.89; sensitivity, 89%; specificity, 68%). A high CAR-HEMATOTOX score resulted in a longer duration of neutropenia (12 vs 5.5 days; P < .001) and a higher incidence of severe thrombocytopenia (87% vs 34%; P < .001) and anemia (96% vs 40%; P < .001). The score implicates bone marrow reserve and inflammation prior to CAR T-cell therapy as key features associated with delayed cytopenia and will be useful for risk-adapted management of hematotoxicity.

Graphical abstract

Figure 1.

Incidence and temporal course of CAR T-cell–mediated hematotoxicity. (A) Cohort description: the primary end point could not be evaluated because of early death (n = 17), incomplete data collection (n = 5), or loss to follow-up before neutrophil recovery and day 60 (n = 1). (B) Proportional incidence of severe anemia (hemoglobin ≤8 g/dL or requiring transfusion), severe thrombocytopenia (platelet count ≤50 g/L), and severe neutropenia (ANC <500 cells per µL; light green) in all patients in the study (n = 235). Neutropenia was further subdivided into protracted (≥7 days) and prolonged (ANC < 1000 cells per µL after day 21). The darker shade of green indicates profound (ANC < 100 cells per µL) neutropenia. (C) Aggregated median ANC over time for 149 patients from the European training and validation cohorts (longitudinal complete blood count sampling was not obtained for the US validation cohort). Measured events per time point are provided in supplemental Table 3. Light shading depicts the 95% confidence intervals (CIs) of the median for each time point. Fludarabine-cyclophosphamide lymphodepletion (LD) was administered on days −5 to −3, and CAR T-cells were transfused on day 0. (D) Aggregated median ANC curves by clinical phenotype of neutrophil recovery. The bar represents the relative distribution of phenotypes. (E-F) Aggregated median platelet count and hemoglobin over time (n = 149). Abs., absolute; LMU, Ludwig Maximilian University; PLT, platelet.

Figure 2.

Markers of impaired hematopoietic reserve, inflammation, and tumor microenvironment are significantly correlated with the duration of neutropenia. Univariate analysis of the influence of the baseline platelet count (A), hemoglobin (Hb) (B), ANC (C), C-reactive protein (CRP) (D), ferritin (E), and BM infiltration (F) on the duration of severe neutropenia (ANC <500 cells per µL) between days 0 and 60 in the training cohort (n = 55). The data concerning BM infiltration was studied via logistic regression analysis; P value is shown for the likelihood ratio test (G²); light shading indicates the 95% asymptotic confidence bands. (A-E) The Spearman correlation coefficient (r) and the respective P values are provided. A positive r value indicates a positive correlation and a negative r value indicates a negative correlation. Light shading indicates the 95% confidence bands of the best-fit line from the simple linear regression. B-NHL, B-cell non-Hodgkin lymphoma.

Figure 3.

The CAR-HEMATOTOX score identifies patients with pronounced myelosuppression after CAR T-cell therapy. (A) Univariate analysis comparing the CAR-HEMATOTOX score to the duration of severe neutropenia in the training cohort (n = 55). The calculated slope (β₁) of the simple linear regression is shown. (B) ROC curve of the influence of the CAR-HEMATOTOX score on the outcome of severe neutropenia at ≥14 days. The AUC and P value are shown. (C) Median duration of severe neutropenia (days 0-60) by CAR-HEMATOTOX score with whiskers indicating the 95% CIs. (D) Relative distribution of clinical phenotypes of neutrophil recovery by CAR-HEMATOTOX score. (E-F) Aggregated median ANC (E) and platelet count (F) over time for patients with either a low or high score. The P value is provided for the comparison of the calculated AUC for high-risk vs low-risk patients (supplemental Figure 6). Mann-Whitney U test ****P < .0001.

Figure 4.

CAR-HEMATOTOX. Determined before lymphodepletion, the score comprises 5 markers of hematotoxicity with additional weighting of the baseline platelet count and ferritin levels. The score discriminates between a high (CAR-HEMATOTOX score ≥2) and low (CAR-HEMATOTOX score 0-1) risk for hematotoxicity.

Figure 5.

The CAR-HEMATOTOX score discriminates between a high risk and a low risk for hematotoxicity in 2 independent patient cohorts. (A-B) Univariate analysis comparing the CAR-HEMATOTOX score to the duration of severe neutropenia in the European (A) and (B) US validation cohorts. (C) ROC curves for the binary outcome of severe neutropenia between ≥4 and <14 days by CAR-HEMATOTOX score for the European and US validation cohorts. (D) Pooled analysis of the median duration of severe neutropenia (days 0-60) by CAR-HEMATOTOX score for both validation cohorts (n = 180); whiskers indicate the 95% CIs. (E) Relative distribution of clinical phenotypes of neutrophil recovery by CAR-HEMATOTOX score for the European validation cohort (n = 80). (F-G) Aggregated median ANC (F) and platelet count (G) over time for patients with a low or high score. The P value is provided for comparison of the calculated AUC for high-risk vs low-risk patients (supplemental Figure 6). Mann-Whitney U test ****P < .0001.

Figure 6.

The impact of the CAR-HEMATOTOX score on concurrent immunotoxicity, clinical outcomes, and the duration of hospitalization. Comparison of (A) CRS and (B) ICANS severity according to American Society for Transplantation and Cellular Therapy (ASTCT) grading among patients with a low or high CAR-HEMATOTOX score. Kaplan-Meier estimates of (C) PFS and (D) OS according to CAR-HEMATOTOX score. (E) Comparison of the overall response rate (ORR) as assessed by positron emission tomography/computed tomography staging on day 90 according to Lugano criteria. In all, 10 of 15 patients with a low CAR-HEMATOTOX score showed evidence of response (complete response [CR], 8; partial response [PR], 2) compared with 9 of 25 patients with high CAR-HEMATOTOX score (CR, 6; PR, 3). (F) Duration of hospitalization from lymphodepletion until discharge or death by CAR-HEMATOTOX score and by severity of hematotoxicity. Severe hematotoxicity was defined as severe neutropenia for ≥14 days between days 0 and 60. Error bars indicate the standard error of the mean (SEM). P-values were determined by Mann-Whitney U test.