The study population was derived from a retrospective single-center cohort of patients who received ICD implantation between 2007 and 2021 for primary or secondary prevention of fatal ventricular arrhythmias. Patients with the following conditions were excluded: 1) those without echocardiography imaging either at baseline or at follow-up, 2) those without echocardiography imaging with adequate quality to be analyzed, 3) those with more than moderate aortic valve stenosis or insufficiency, and 4) those without electrocardiogram (ECG) with adequate quality for measuring. All patients received optimal medical therapy according to the guidelines.

Of the total population, we defined the event group as those who had appropriate ICD shock or anti-tachycardial pacing (ATP) for fatal ventricular arrhythmias, i.e., ventricular tachycardia (VT) or ventricular fibrillation (VF), during the follow-up after ICD implantation. The control group was defined as those who did not receive any appropriate ICD therapy during the follow-up.

We obtained information on demographic features, purpose, and indication of ICD implantation, and the occurrence of fatal arrhythmic events by retrospectively reviewing the electronic medical records. The data on medications and echocardiography images were acquired at two time points: at baseline and follow-up. The time interval between baseline and follow-up echocardiographic exam was pre-determined for asymptomatic participants, while was adjusted according to the patients’ condition and physicians’ discretion. The “follow-up” time-point in this study indicated the nearest point before the fatal arrhythmic event for the event group and the latest follow-up exam for the control group.

All participants had a pair of transthoracic echocardiography (TTE) image sets. TTE was performed according to the recommendations of current guidelines, using the commercially available devices (Vivid E9 or E95, GE Medical Systems; EPIQ 7C or iE33, Philips Healthcare; ACUSON SC2000 or SEQUOIA, Siemens Medical Solutions). The left ventricular ejection fraction (LVEF) was measured by the biplane Simpson's method using the apical 4- and 2-chamber views. The apical 5-chamber view was necessarily acquired for the measurement of EMW.

The QTc was calculated by the tangent method and corrected using the Bazett formula. An ECG embedded in the echocardiographic image for EMW measurement was also used to calculate the QTc. The continuous-wave doppler images on the apical 5-chamber view were used for EMW measurement. The EMW was calculated as the time interval between the onset of QRS complex and aortic valve closure (QAoC) minus the QT interval from the ECG of the same beat and measured in the continuous wave doppler image on the apical 5-chamber view of TTE (Fig. 1). We exported all echocardiography images into the DICOM format and loaded them into the dedicated analysis software (RadiAnt DICOM Viewer, Version 2022.1.1). Using the measurement tool incorporated in the software, we could accurately measure the QAoC duration and QT interval in millisecond units. This approach enabled us to consistently measure EMW, regardless of sweep speed or image source from various vendors. In the case of atrial arrhythmias such as atrial fibrillation (AF), atrial flutter (AFL), or other conditions with expected large beat-to-beat variability, we tried to measure the parameters for EMW based on the average values across multiple beats as much as possible (optimally at least 3 beats). Since PVCs are known to significantly impact the measurement of EMW, we excluded cases where it was not possible to measure EMW appropriately due to PVC beats. A primary investigator (T.M.R.) who was blinded to the individual patient's information measured QTc and EMW objectively. The measurements of both parameters were repeated on the same beat of randomly selected subsets of the study population by the primary observer to evaluate the intra-observer variability, and by an independent investigator (H.J.A.) to assess the inter-observer variability.

Data were presented as numbers and frequencies for categorical variables, and as mean ± standard deviation or median (1^st quartile–3^rd quartile) for continuous variables. We conducted a Shapiro-Wilk normality test to verify the normality distribution of the continuous variables. For comparisons between groups, the chi-square test or Fisher’s exact test was used for categorical variables, and an unpaired Student’s t-test or Mann-Whitney test was used for continuous variables, as appropriate. Descriptive statistics and density distribution of EMW were compared between the event and control groups. The correlations of QTc and EMW were displayed in the scatter plots and the Pearson’s correlation coefficients were calculated in both the event and control groups. Univariable and multivariable logistic regression analyses were performed to investigate which parameter is independently associated with the occurrence of a fatal arrhythmic event, by adjusting for potential confounders including secondary prevention, primary electrical disorder, hypertrophic cardiomyopathy, LVEF, and use of class III antiarrhythmics at baseline. Covariates used in the multivariable analysis were selected if they were significantly associated with a fatal arrhythmic event in the univariable logistic regression analysis (P value < 0.2). The discriminant ability of EMW for predicting the occurrence of a fatal ventricular arrhythmic event was evaluated using the receiver operating characteristic (ROC) curve comparing with a multivariable logistic regression model including clinical variables. Best cutoff value of EMW was calculated from the multivariable logistic regression model using the Liu method.7 EMW was also compared between the event and control group in various exploratory subgroups. The intraclass and interclass correlation coefficients (ICCs) were calculated to evaluate the reliability of EMW measurements. A P value of < 0.05 was considered statistically significant. All analyses were conducted using Stata software, version 14.0 (StataCorp., College Station, TX, USA).

The study protocol was approved by the Institutional Review Board of Seoul National University Hospital (2104-007-1208) and conducted according to the principles of the Declaration of Helsinki. Informed consent was waived because of the retrospective nature of the study.

The baseline characteristics of the overall population are listed in Table 1. Of 245 patients (mean age 67.2 years, 63.7% men), 20.0% (49 patients) experienced VT or VF episode that was treated with appropriate ICD shock or ATP. Secondary prevention was considerably more prevalent in the event group. The event group had a lower prevalence of cardiac resynchronization therapy with defibrillator implantation than the control group. QTc at follow-up (QTc-FU) was substantially longer in the event group than in the control group at follow-up, while was comparable between the two groups at baseline (QTc-base). While both groups had similar heart rates at baseline and follow-up, the event group had a greater negative EMW at both baseline (EMW-Baseline) and follow-up (EMW-FU) (Fig. 2). The median duration between baseline and follow-up was 2.4 (1.0–4.8) years.

The scatter plot depicts the relationship between QTc and EMW in the event and control groups (Fig. 3). QTc and EMW had significant weak to moderate negative correlations in the event group. The correlation was highest between QTc-FU and EMW-FU in the event group. (r = −0.542, P < 0.001). QTc fluctuated widely within the given range of EMW, demonstrating a relatively weak association between QTc and EMW especially in the control group.

ICD implanted for secondary prevention, hypertrophic cardiomyopathy, LVEF, AF or AFL at baseline, use of class III antiarrhythmics at baseline, QTc-FU, EMW-Baseline, and EMW-FU were found as possible predictors of events and included in a multivariable logistic regression model (Table 2). Among them, secondary prevention, QTc-FU, EMW-Baseline, and EMW-FU remained significant predictors of fatal arrhythmic events. Decrease of 1 ms in EMW-Baseline and EMW-FU were linked with a 2% and 6% increase in the probability of events, respectively.

Including EMW-Baseline as an additional covariate significantly improved the predictive ability of a multivariable model including secondary prevention, hypertrophic cardiomyopathy, primary electrical disorder, LVEF, use of class III antiarrhythmics at baseline, and AF or AFL at baseline (area under the curve [AUC] 0.77 [0.70–0.84] vs. AUC 0.72 [0.64–0.80], P = 0.004 against model with clinical variables). The predictive ability of EMW-FU alone (AUC 0.87 [0.81–0.94]) considerably surpassed that of other multivariable models; P = 0.060 against model with clinical variables, P = 0.030 against model with clinical variables and EMW-Baseline (Fig. 4). The cutoff value of EMW-FU for the best performance of model was −4.0 ms with sensitivity of 88% and specificity of 82% (Fig. 5).

The results were consistent across the majority of exploratory subgroups (Table 3). The negativity of EMW-FU in the event group was prominent in patients with non-ICM (−40.0 [−67.0 to −20.0 ms]) and non-ICM due to valvular heart disease (−61.0 [−80.0 to −31.0 ms]). EMW-Baseline was also significantly more negative in the event group in or for individuals who had non-ICM (especially DCM) or who had implanted ICD for secondary prevention. The performance of predictive models were compared in subgroups (Table 4). Among all subgroups, the model with EMW-FU alone showed the best predictive ability with the AUC of ROC ranging from 0.83 to 0.96. The predictive ability of clinical variables and EMW-Baseline was better in cardiomyopathies (both ICM and non-ICM) than in primary electrical disorder.

To enhance the reproducibility of our study, a sensitivity analysis was conducted on a homogeneous patient population of 217 individuals who experienced only VF events excluding monomorphic VT cases, and 171 patients who did not undergo CRT. The results were consistent, revealing a significant negativity of EMW-FU (Supplementary Table 1).

Intra-observer variability was excellent for EMW measurement (ICC 0.98, [0.96–0.99]). Inter-observer variability between two independent investigators was also excellent for EMW measurement (ICC 0.96 [0.92–0.98]).