Combined prediction model for disease-free survival after allo-HSCT as identified by regression tree analysis. (A) Combined prediction model for disease-free survival in our cohort as identified by regression tree analysis. Numbers in circles represent the number of patients at that node. Split to the left represents higher-risk criterion. Numbers in boxes represent the number of relapses that were predicted by the split variable in the respective risk group. (B) Subsequently, the impact of the identified splits (bone marrow infiltration > 50%, and the Ang-2 cut-off point of 2.28 ng/mL) on disease-free survival were analyzed. Overall comparison (Mantel-Cox) between the 3 risk groups was significant (P < .001). In addition, pair-wise log-rank test was calculated for the 2 split points in a consecutive fashion and confirmed the independent statistical significance between each of the 3 risk groups (bone marrow infiltration ≥ 50% vs < 50%, P < .001, n = 90; Ang-2 > 2.28 vs < 2.28 ng/mL, P = .049, n = 71). The corresponding Kaplan-Meier curves of all 3 clearly distinguishable risk groups are shown in Figure 6.