Fig. 3: Random survival forest model for impact of clinical and/or genetic variables as well as pairwise interactions on prognosis.

Prognostic and potential predictive impact of 18 most relevant clinical and/or genetic variables (A, B) and pairwise interactions (C, D) in 475 patients on overall and event-free survival using random survival forests. The prognostic impact of a variable is measured via VIMP (variable importance). Higher VIMP values indicate that a variable may have prognostic impact on the survival endpoint. EFS event-free survival, OS overall survival.