TABLE 2

Prognostic Potential of Different Models Trained in Training Cohort and Evaluated in Validation and Merged Groups

ModelTraining (n = 15)Validation (n = 22)Merged (n = 37)
MFMISO/TCP
 Concordance0.77 (0.67–0.82)0.71 (0.64–0.77)0.74 (0.71–0.77)
 P0.001*0.004*<0.0001*
 HR17.0 (1.9–143)6.7 (1.5–30)9.4 (2.9–31)
rHV
 Concordance0.57 (0.36–0.57)0.63 (0.58–0.67)0.61 (0.57–0.67)
 P0.2690.001*0.020
 HR8.6 (0.22–331)
TMRmax
 Concordance0.58 (0.40–0.62)0.60 (0.47–0.61)0.61 (0.54–0.68)
 P0.4350.0280.043
 HR
GTV
 Concordance0.67 (0.62–0.77)0.530.60 (0.52–0.66)
 P0.0830.3170.031
 HR
LogitTMR-V
 Concordance0.73 (0.70–0.85)0.58 (0.53–0.62)0.65 (0.62–0.70)
 P0.010*0.10.001*
 HR2.4 (0.56–11)3.1 (0.91–11)
  • * P < 0.01.

  • Confidence interval estimation not possible in this case.

  • rHV = relative hypoxic volume; logitTMR-V = logistic regression model trained for TMRmax and GTV.

  • Concordance index is for univariable Cox model. P values are for log-rank test regarding Kaplan–Meier analysis. Data in parentheses are 95% confidence intervals.