Prognostic risk models of patients with or without HSCT. (A) Reducing the effect of multicollinearity by compressing regression coefficients of variables in the model to 0 for patients undergoing HSCT. (B) The λ.1se was 1 standard error of λ, which corresponded to the smallest partial likelihood deviance of evaluation index for patients with HSCT. (C) Receiver operating characteristic (ROC) curves for sCD25 and EBV-DNA load of patients undergoing HSCT. (D) Nomogram used to predict time-related morality in patients with HSCT. (E) Reducing the effect of multicollinearity by compressing regression coefficients of variables in the model to 0 for patients without HSCT. (F) The λ.min corresponded to the smallest partial likelihood deviance of evaluation index for patients without HSCT. (G) ROC curves for indicators of patients without HSCT. (H) Nomogram used to predict time-related morality in patients without HSCT.