Fig. 3

Application of single-sample predictors (SSPs) to stratify patients for treatment. The analysis was performed with data stratified for estrogen receptor (ER) status. SSPs developed in radiotherapy (RT)-untreated patients (RT-) were used to estimate the risk without giving RT. If they were predicted as having low risk of IBTR without RT, they were assigned to the “No-RT” groups. If predicted as high risk without RT, a SSP developed in RT+ tumors was applied. If predicted as having low risk with RT, they were assigned to the group “Give RT” and if predicted as having high risk with RT, they were assigned to the “More-treatment” group. The difference in risk of ipsilateral breast tumor recurrence with or without RT was visualized using the Kaplan-Meier method and tested with the log-rank test for ER+ tumors (a) and ER- tumors (b). Among ER- tumors (b), only two were RT- and we thus analyzed the prognostic effect of the groups assigned