Decision tree ML model predicts early relapse within 6 months after axi-cel using age and 6 first laboratory values within 24 hours immediately after axi-cel infusion. (A) The decision tree ML model developed to identify patients with higher risks of experiencing early relapse after receiving axi-cel. (B) The ROC curves for each leave-1-group-out validation and OOS testing data set. (C) The 7 patient-level predictors important to the decision tree ML model’s predictions. Kaplan-Meier estimates of (D) PFS and (E) OS of groups A and B predicted by the ML model in OOS testing data set. Groups A and B represent subgroups of patients without or with risk of experiencing relapse within 6 months after receiving axi-cel, respectively. The x-axis shows days since administration of axi-cel. (F) Decision curve analysis of the ML model across all threshold probabilities and (G) between 0 and 0.4. AUC, area under the curve; CV, crossvalidation; PT, prothrombin time; ROC, receiver operating characteristic.