Fig. 3

miR-548c-5p as factor in relapse prediction model. Performances of two models are compared to measure the improvement of relapse prediction by the inclusion of the miR-548c-5p as a 4th variable, with the first three variables being node invasion, CK5/6 expression, and tumor size. a Comparison of ROC curves computed with the relapse probability calculated by the model including miR-548c-5p (solid line) and without miR-548c-5p (dash line). b Patients were classified in two groups: high and low risk of relapse, according to the threshold needed to obtain 90 % sensitivity in relapse prediction. Comparison of Kaplan-Meier curves computed with the patient group affectation calculated by the model including miR-548c-5p (solid line) and without miR-548c-5p (dash line). c Classification performances of the two models at thresholds fixed to obtain 90 % sensitivity in relapse detection. d Coefficient and odds ratio of the model including miR-548c-5p and e without miR-548c-5p. f Comparative expression levels of miR-210, miR-503-5p and BRCA1 mRNA in the patients with <10 % probability of relapse (remission) and >90 % probability of relapse (relapse). These probabilities were calculated by the prediction model including miR-548c-5p