A Predictive Model of Severe Cytokine Release Syndrome After Coadministration of CD19- and CD22-Chimeric Antigen Receptor T-Cell Therapy in Children With B-Cell Hematological Malignancies Based on Patient-Reported Outcomes

Abstract

Background Chimeric antigen receptor T-cell therapy–related severe cytokine release syndrome (sCRS) has seriously affected the life safety of patients. Objective To explore the influencing factors of sCRS in children with B-cell hematological malignancies and build a risk prediction model. Methods The study recruited 115 children with B-cell hematological malignancies who received CD19- and CD22-targeted chimeric antigen receptor T-cell therapy. A nomogram model was established based on symptomatic adverse events and highly accessible clinical variables. The model discrimination was evaluated by the area under the receiver operating characteristic curve. The calibration of our model was evaluated by the calibration curve and Hosmer-Lemeshow test. The bootstrap self-sampling method was used to internally validate. Results Thirty-seven percent of the children experienced sCRS. Indicators included in the nomogram were tumor burden before treatment, thrombocytopenia before pretreatment, and the mean value of generalized muscle weakness and headache scores. The results showed that the area under the receiver operating characteristic curve was 0.841, and the calibration curve showed that the probability of sCRS predicted by the nomogram was in good agreement with the actual probability of sCRS. The Hosmer-Lemeshow test indicated that the model fit the data well (χ 2 = 5.759, P = .674). The concordance index (C-index) obtained by internal validation was 0.841 (0.770, 0.912). Conclusions The nomogram model constructed has a good degree of discrimination and calibration, which provides a more convenient and visual evaluation tool for identifying the sCRS. Implications for Practice Incorporation of patient-reported outcomes into risk prediction models enables early identification of sCRS.