The success of cardiac resynchronization therapy in selected patients with heart failure and reduced left ventricular (LV) ejection fraction not adequately treated with medications has made it more important to refine the criteria for diagnosing LV dyssynchrony. Traditional criteria for LV dyssynchrony have focused on electrical criteria—e.g., the presence of a left bundle-branch block (LBBB) and a QRS duration >150 msec. More recently, mechanical indices detected by imaging techniques such as gated single photon emission computed tomography imaging and echocardiography (echo) have been utilized as additional parameters to refine the diagnosis of LV dyssynchrony.
In this context and with the aim of identifying the key determinants of LV dyssynchrony, Kong and coauthors studied the interaction and contribution of mechanical and electrical dyssynchrony parameters in 412 chronic heart failure patients using machine learning and explainable artificial intelligence techniques—specifically, a random forest model and SHapley Additive exPlanations analysis. These heart failure patients were divided into two groups: 239 with synchronous LV contraction (SG) and 173 patients with asynchronous LV contraction (AG). SG was defined as an electrocardiographic QRS duration ≤120 msec, echo LV end-diastolic volume ≤150 ml, and standard deviation of time-to- peak longitudinal strain in 18 LV segments ≤33 msec by speckle tracking echo. The authors found that compared with the SG group, AG patients had significantly longer QRS duration, higher NT-proBNP levels, and higher LV end-diastolic and LV end-systolic volumes. Their machine learning analysis identified LV ejection fraction, LV end-systolic volume, and LV posterior wall dyssynchrony by strain—specifically in the posterior basal, apical, and midsegments—as the primary predictors of LV dyssynchrony, superior to the electrical markers. The presence of LBBB increased the likelihood of LV dyssynchrony by 3.2-fold. The integrated electrical-mechanical model studied by the authors showed excellent performance and accuracy, outperforming traditional dyssynchrony indices such as QRS prolongation and LBBB. It is of interest that the authors’ findings of predictors of mechanical LV dyssynchrony detected by speckle tracking echo in the posterior LV wall segments have been previously reported using single photon emission computed tomography myocardial perfusion imaging with summed rest and stress scores.
There were a number of limitations of the study, most of which were cited by the authors—including the fact that their study was a single-center study with apparently limited ethnic diversity. In addition, they did not include follow-up studies in their heart failure patients to investigate whether their combined model using mechanical and electrical abnormalities was more powerful than LBBB or prolonged QRS duration in predicting successful cardiac resynchronization or other therapy. In this regard, Obaid et al have previously demonstrated in 103 heart failure patients that antero-septal to posterior intraventricular delay and NT-proBNP levels were independent predictors for cardiac events over a 1-year follow-up period. Furthermore, it is perhaps not surprising, since the definition for LV dyssynchrony that the authors used for patient stratification included echo findings—e.g., LV ejection fraction, LV volumes, and time-to-peak longitudinal strain in 18 LV segments—that the authors found that mechanical factors detected by echo were more important than traditional electrocardiographic factors used to define LV dyssynchrony. Nonetheless, the authors’ findings contribute to the important evolving literature attempting to further refine the definition of cardiac dyssynchrony so as to provide, as the authors note, “actionable thresholds for refining patient selection for cardiac resynchronization therapy” and potentially other cardiac therapies.
CRediT authorship contribution statement
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
