Patients undergoing cardiac resynchronisation therapy are heterogeneous. Therefore, the prediction of CRT-response and outcome is challenging. The aim of this study was to evaluate the relative impact of clinical, electrocardiographic, and echocardiographic variables on the left ventricular (LV) remodelling and prognosis of CRT-candidates by the application of machine-learning (ML) approaches.

In total, 193 patients undergoing CRT (mean age 67±11 years, QRS width: 167±21 msec) were prospectively included in this multicentre study. We used a combination of the Boruta algorithm and random forest methods (RF) to identify features predicting both CRT-volumetric response and prognosis. The model performance was tested by the area under the receiver-operating curve (AUC). We also applied the K-medoid method to identify clusters of phenotipically-similar patients.

From 28 clinical, electrocardiographic, and echocardiographic-derived variables, 16 features were predictive of CRT-response, and 11 features were predictive of prognosis. Among the predictors of CRT-response, 6 variables (38%) pertained to right ventricular (RV) size or function. The tricuspid annular plane systolic excursion was the main feature associated with prognosis. The selected features were associated with a very good prediction of both CRT-response (AUC: 0.81, 95% CI: 0.74–0.87) and outcome (AUC 0.84, 95% CI: 0.75–0.93) (Fig. 1A–B). An unsupervised ML approach allowed the identifications of two phenogroups of patients who differed significantly in clinical parameters, biventricular size and RV function and had significantly different prognosis (log-rank P<0.0001; HR: 4.7, 95% CI: 2.1–10.0, P<0.001) (Table 1).

ML can reliably identify clinical and echocardiographic features associated with CRT-response and prognosis. Our results underscore the value of the assessment of RV-size and function parameters for the risk stratification of CRT-candidates.