We tested the hypothesis that a machine learning (ML) algorithm utilizing both complex echocardiographic data and clinical parameters could be used to phenogroup a heart failure (HF) cohort and identify patients with beneficial response to cardiac resynchronization therapy (CRT).

No abstract

Background-Despite benefits of cardiac resynchronization therapy (CRT) in patients with severe but less symptomatic heart failure, approximately 30% of patients do not fully respond to treatment. We hypothesized that a combined assessment of left ventricular (LV) dyssynchrony and contractile function by strain-based imaging would identify patients who would most benefit from CRT. Methods and Results-We studied 1077 patients with New York Heart Association class I/II, LV ejection fraction Յ30%and QRS width Ն130 ms enrolled in the Multicenter Automatic Defibrillator Implantation Trial-Cardiac Resynchronization Therapy trial with sufficient echocardiographic image quality for cardiac deformation analysis (implantable cardioverter-defibrillator [ICD], nϭ416; CRT, nϭ661). Patients were assigned to CRT plus an ICD or to ICD alone in 3:2 random assignment. We assessed the degree to which baseline echocardiographic assessments of dyssynchrony, measured as the standard deviation of time-to-peak transverse strain over 12 segments, contractile function, measured as global longitudinal strain, or both predicted the effect of treatment on the primary outcome of death or heart failure. With 213 primary events occurring over a mean of 2.4 years, the benefit of CRT plus an ICD relative to ICD alone was greatest in patients with mild to moderate dyssynchrony (time-to-peak transverse strain standard deviation, 142 to 230 ms) and greater baseline contractile function (global longitudinal strain ՅϪ8.7%). Overall, those patients with mild to moderate dyssynchrony and those with best contractile function at baseline demonstrated the greatest benefit from CRT (adjusted hazards ratio, 0.20; 95% confidence interval, 0.09 to 0.44). Dyssynchrony and global longitudinal strain predicted response to CRT independent of each other, QRS width, LV ejection fraction, and presence versus absence of left bundle-branch block, although the observed benefit remained greatest in patients with left bundle-branch block. Conclusions-Both

Cardiac resynchronization therapy resulted in a significant improvement in both LV dyssynchrony and contractile function measured by GLS compared with ICD only and these improvements were associated with better subsequent outcomes.

Baseline LV dyssynchrony did not predict VT/VF/death or VT/VF in mild heart failure patients with or without LBBB. CRT-induced improvement of LV dyssynchrony was associated with significant reduction of ventricular arrhythmias in patients with LBBB.