Figure 3.

The classification trees showing the prognostic signature consisting of 2 variables, MTV and MH, whose dichotomization was significantly correlated with survival. MTV alone, when its value is low, identifies the prognostic category node 2 with best survival probability. To separate the prognostic categories node 4, with intermediate survival rates, from node 5, with poorer survival, the value of MTV must be high (node 3), and that of MH must be low or high, respectively. The Kaplan-Meier curves show statistically different overall and progression-free survival, either in both the testing and the validation patient sets. For each inner node, the Bonferroni-adjusted P values are given, which take into account the multiple testing and may differ from the P values of the log-rank test in the univariable analysis (Table 3).