Fig. 7

Computational analysis for prediction of time to relapse (TTR).a Cross-validated Harrel’s C-index using random survival forest models. The variables are selected by importance using minimal depth. b Minimal depth ranking of covariates. c-d Effect of the covariate in metastasis dissemination according to the value of the covariate. The histogram corresponds to the covariate data and the vertical bars are the corresponding values of the metastasis dissemination parameter (μ) distribution according to the value of the covariate. (See Ref. A8 for more details). c CFB, specific parameter values are b = 1.04 (Relative Standard Error (RSE) = 14.90%), c = 0.22 (RSE = 24.26%) and dif = − 0.67 (RSE = 11.44%) d Saa2, specific parameter values are b = 0.32 (RSE = 32.82%), c = (RSE = 39.47%) and dif = − 0.89 (RSE = 61.98%). e Goodness-of-fit for the model with the effect of CFB and the data at different thresholds. f Goodness-of-fit for the model with the effect of SAA2 and the data at different thresholds. g-h Individual predictions of two patients, with probabilities to have metastasis at diagnosis or not to have metastasis after 5 years. The plot corresponds to the predicted DMFS curve for the individual patients, which allows to calculate the predicted TTR