Fig. 2.

Decision tree machine learning approach. A Decision trees calculated using the Classification and Regression Trees (CART) algorithm in the whole study sample. Predictors considered in the analysis were age, treatment group allocation (ASV or control), male sex, SBP < 120 mmHg, diabetes, diuretics, cardiac device and 6 min walk distance, NT‐proBNP, atrial fibrillation, in addition to the microRNA candidate: miR-133a-3p. The results are presented in a binary decision tree that was constructed by splitting a node into two child nodes repeatedly. Generation of novel nodes was based on the selected predictors and cutoffs. Incidence rates (IR) of events per 100 patients/year, number of patients per node and hazard ratios (HR) for the eight final nodes defined by the regression tree model including microRNAs are included. The length of each color in the bands is proportional to the percentage of the total time that patients are submitted to the risk range; B Kaplan–Meier curves illustrated differences among nodes in the observed time-to-event outcome. Patients at risk for each subgroup of patients identified are displayed; C Incremental area under the cumulative/dynamic ROC curve (iAUC) of the ordinal risk of the final nodes. MicroRNA expression profiles were assessed using RT-qPCR. Relative quantification was performed using miR-486-5p for normalization. Relative expression levels were log-transformed for statistical analyses