Generation of the derivation and validation cohorts and identification of prognostic factors in the derivation cohort
The whole cohort was divided in a 2:1 ratio into a derivation and validation cohort in a randomized fashion. Univariable Cox regression analyses identified factors prognostic for EFS in the derivation cohort. Factors with p<0.05 on univariable analyses were included for multivariable analysis in a backward stepwise fashion based on likelihood ratio. Factors with p<0.05 in the final multivariable model in the derivation cohort were used to formulate the risk score.