Demographic characteristics, medication history and laboratory tests were collected from electronic medical records on admission. Two-dimensional echocardiographic examination was performed 3 days before ICD insertion by experienced sonographers and interpreted by well-trained cardiologists, using commercially available equipment. Sequential cardiac cycles were recorded during breath holding with stable electrocardiography tracing. Chamber dimensions and functional parameters were measured according to the American Society of Echocardiography and the European Association of Cardiovascular Imaging [12]. Right ventricular outflow tract diameter was measured from the anterior RV wall to the interventricular septal-aortic junction in a standard parasternal long-axis (PSLAX) end-diastole view, as depicted in Supplementary Fig. 1 [13].

The follow-up period began on the first day after implantation. Device interrogation contained a review of the stored intracardiac electrograms 3 months later and every six to twelve months. The primary endpoint was the first appropriate shock triggered by ICD-monitored life-threatening ventricular tachycardia (VT)/ventricular fibrillation (VF) and undertreated SCD adjudicated by association-certified electrophysiologists. Shocks were determined appropriate if the preceding rhythm was classified as VT/VF. Inappropriate therapies and antitachycardia pacing (ATP) were excluded from the initial outcome. The secondary endpoint was non-arrhythmic mortality, defined as a composite of death or cardiac transplantation without exposure to any sustained VT/VF during follow-up. The survival status was obtained from medical health records or telephone calls until February 2022. The dates for the censoring of interrogation and death are not necessarily the same.

Firstly, to explore an additional value of RVOTD based on existing VT/VF risk tools, the validated Seattle Proportional Risk Model (SPRM) was used as one of the basic models (male sex, younger age, no diabetes, lower left ventricular ejection fraction, lower systolic blood pressure, lower creatinine level, lower serum sodium level, better New York Heart Association functional class, higher body mass index and digoxin use) [14]. Another basic model is the Multicenter Automatic Defibrillator Implantation Trial (MADIT)-ICD benefit score, consisting of VT/VF score (LVEF $\le$25%, atrial arrhythmia, heart rate $>$75 bpm, systolic blood pressure 140 mmHg, myocardial infarction, age 75 years, male, prior non-sustained VT) and non-arrhythmic mortality score (New York Heart Association $\ge$II, diabetes, body mass index 23 kg/m${}^2$, atrial arrhythmia, LVEF $\le$25%, and age $\ge$75 years) [15]. Cardiac resynchronization therapy-cardioverter defibrillator (CRT-D) recipients were not included considering that varying response rates of cardiac resynchronization therapy would confound their future effect on outcomes. Variable types and thresholds for categorization of numeric variables included in current analysis were basically the same as the original ones, and some corrections and explanations are stated as follows. Serum sodium values were analyzed as continuous units below 145 mEq/L, which is the upper limit of reference value in our hospital labs. When calculating MADIT-ICD scores, the points assigned for each variable were consistent with the original ones, too. All prior VT/VF events were analyzed when measuring MADIT-ICD VT/VF score.

Secondly, a new risk prediction model for ICD benefits incorporating RVOTD was introduced. We identified factors associated with increased risk for VT/VF, after accounting for non-arrhythmic mortality as a competing risk, and created a VT/VF risk score. Then, we used a similar method to construct the non-arrhythmic mortality risk score for death without a prior VT/VF as the endpoint. Next, we allocated each individual into a risk stratum by calculating a newly developed RVOTD-ICD benefit score that combined the VT/VF risk score and the non-arrhythmic mortality risk score for both outcomes. The whole population was separated into three benefit groups: (i) Highest benefit (highest VT/VF risk and lowest non-arrhythmic mortality risk), (ii) Intermediate benefit (higher VT/VF risk and lower non-arrhythmic mortality risk), and (iii) Lowest benefit (lowest risk of VT/VF and highest risk of non-arrhythmic mortality). Finally, we compared the RVOTD-ICD benefit score with MADIT-ICD benefit score by evaluating the overall survival benefit among these populations.

Continuous data were presented as the median and the interquartile range (IQR) or mean and standard deviation. Categorical data were expressed as frequencies with percentages. Baseline characteristics of the lower and upper RVOTD patients were compared using $\chi$${}^2$ test for categorical variables, the unpaired t-test for normally distributed continuous variables, and Kruskal‒Wallis test for continuous variables with nonnormal distribution.

Step 1—selection of RVOTD and other prognostic factors.

We performed analyses of the unadjusted cumulative incidence rates for VT/VF events and non-arrhythmic death, illustrated by Kaplan–Meier curves. Then, a multivariable Fine-Gray model, using non-arrhythmic mortality as a competitive risk and VT/VF as the endpoint, was used to adjust the effect of RVOTD in the final model. Variables in Seattle Proportional Risk Model, the MADIT-ICD VT/VF score and Non-arrhythmic Mortality Score, as well as other potential risk factors available for the entire cohort were candidate variables, whose p values less than 0.05 in univariable analysis were then included in the multivariable model to test for an independent association between outcomes and RVOTD. Backward selection was used based on Akaike Information Criterion (AIC) rule. The significance level for staying in the final model is 0.05. The same stepwise selective Cox regression for non-arrhythmic mortality was performed as for VT/VF. The follow-up time was calculated as the time between ICD implantation and the outcome events or censoring. Furthermore, RVOTD was added with all variables in SPRM and MADIT-ICD scores to assess the incremental value based on existing risk models. Interaction analysis were performed to examine potential heterogeneity of RVOTD between subgroups.

Step 2—development and validation of RVOTD-ICD benefit score.

Variables related to either VT/VF or non-arrhythmic death were selected based on step 1. To create a simple scoring method, numeric variables were categorized by the use of cut-off points. The age range was categorized by ten years. Log NT-proBNP was categorized by the quartile distribution. Thresholds for categorization of echocardiographic parameters were cut off by median. Each variable was then assigned a numeric value based on the relative value of its regression coefficient in the multivariate regression model. The prediction scores for VT/VF and non-arrhythmic mortality were separately validated by measuring discrimination using time-dependent receiver operating characteristic curve with 1000 randomly bootstrapped samples. Calibrations were assessed by comparing observed risk with the predictive risk of two RVOTD-ICD scores.

Given the VT/VF rate was approximately three times as non-arrhythmic death rate in the whole population, and in hope of creating positive scores, RVOTD-ICD benefit score was calculated as follows: 3 $\times$ VT/VF score – non-arrhythmic death score + 50, in which higher score denotes a higher long-term benefit from ICD implantation. The cohort was trichotomized into three groups based on patient-specific risk for VT/VF and non-arrhythmic mortality measured as ICD benefit score. Within each group, we used cumulative incidence function curves to illustrate both outcomes.

Step 3—comparison with MADIT-ICD benefit score.

ROC curve analysis using nonparametric estimates of the area under the curve (AUC) was performed to compare the predictability of VT/VF score and non-arrhythmic death score between MADIT-ICD and RVOTD-ICD models. Patient discrimination and reclassification were also evaluated using continuous net reclassification improvement (cNRI). Finally, the clinical usefulness and net benefit of both benefit score systems were estimated with decision curve analysis to test whether the new model would stratify long-term risk of whole ICD population in different ranges of threshold probabilities.

SPSS version 26.0 (IBM Corp., Armonk, NY, USA) and R version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria) were used for statistical analyses. Missing data were handled by multiple imputations. A two-sided p-value 0.05 was considered statistically significant unless specified otherwise.

The 954 total ICD recipients were divided according to median RVOTD (23 mm) (Fig. 1). The population was on average 58.8 years old, and predominantly male (79.0%). The ICD manufacturers included Medtronic, Abbott, Biotronik, and Boston Scientific. A total of 604 patients underwent implantation of a single-chamber ICD. Supplementary Fig. 2 presents the distribution of RVOTD in the cohort. The baseline characteristics of the patients are shown in Table 1. Patients with a higher median RVOTD had a higher body mass index. They were more likely to suffer from atrial fibrillation and less likely to suffer from coronary arterial disease. In terms of echocardiographic parameters, they had higher left atrial diameters, but the left ventricular end-diastolic diameter and LVEF were similar. Additionally, they were more likely to be prescribed digoxin, and their erythrocyte sedimentation rates and lactic dehydrogenase concentrations were higher.

During a median interrogation follow-up of 2.83 years (interquartile range: 1.33–5.27 years) and median death follow-up of 3.85 years (interquartile range: 2.14–6.37 years), 285 patients experienced appropriate ICD shock (29.9%) or SCD (0.2%), and 140 patients died without experiencing any sustained VT/VF since implantation (14.7%). In the unadjusted time-to-event curves, patients with a higher median RVOTD had a higher cumulative incidence of VT/VF events than their lower counterparts (Fig. 2A, Table 2) (hazard ratio [HR], 1.48; 95% confidence interval [CI], 1.17–1.87; p = 0.001). However, the non-arrhythmic mortality risk was not significantly different between the two groups (Fig. 2B, Table 2; HR, 1.09; 95% CI, 0.78–1.52; p = 0.628). A multivariable model fully adjusted for all possible confounders in the univariate analysis (Supplementary Table 1) showed consistent adverse effects of RVOTD in VT/VF events (HR per standard deviation (SD): 1.22; 95% CI, 1.11–1.33; p = 0.002). Receiver operating characteristic curve analysis confirmed the predictive value of RVOTD for VT/VF events (area under the curve [AUC], 0.61; 95% CI, 0.55–0.66; p 0.001), with the best cutoff value at 27 mm. The fully adjusted Seattle Proportional Risk Model and Multicenter Automatic Defibrillator Implantation Trial–Implantable Cardioverter-Defibrillator (MADIT-ICD) models also suggested that RVOTD was a robust indicator of VT/VT events, both as categorical and continuous variables (Table 2).

Fig. 2.

Incidence of VT/VF (A) and non-arrhythmic mortality (B) in patients with lower or higher right ventricular outflow tract diameters. RVOTD, right ventricular outflow tract diameter; VT, ventricular tachycardia; VF, ventricular fibrillation.

Subgroup analyses demonstrated that the link between RVOTD and VT/VF was homogenous across various subgroups of patients, particularly among patients with or without left ventricle hypertrophy or different LVEF groups (Supplementary Fig. 3). Interestingly, the risk of RVOTD may be even higher in those who were not treated with guideline-directed pharmacotherapy, including renin-angiotensin-aldosterone system (RAAS) inhibitors, $\beta$-blockers, and mineralocorticoid receptor antagonists (p for interaction 0.05). A sensitivity analysis, including monitoring ATP as the primary endpoint, revealed a robust association between RVOTD and ventricular tachyarrhythmia (Supplementary Table 2).

Together with a higher RVOTD, younger age, diuretic use, prior non-sustained VT (NSVT), and prior sustained VT/VF were eventually identified as having an increased risk of VT/VF in the Fine-Gray regression model (Table 3). The AUC of the RVOTD-ICD VT/VF score of 1, 2, and 3 years were 0.64 (95% CI, 0.59–0.68), 0.64 (95% CI, 0.59–0.68), and 0.62 (95% CI, 0.57–0.66), respectively (Supplementary Table 3). On the other hand, six factors were identified as predictors of non-arrhythmic mortality: older age, diabetes, higher left ventricular end-diastolic diameter (LVEDD), New York Heart Association (NYHA) class III/IV and higher NT-proBNP quartile were associated with increased risk of non-arrhythmic mortality, whereas RAAS inhibitors use was related to reduced risk (Table 3). The AUC of the RVOTD-ICD non-mortality score of 1, 2, and 3 years were 0.81 (95% CI, 0.73–0.88), 0.75 (95% CI, 0.69–0.81), and 0.78 (95% CI, 0.73–0.83), respectively (Supplementary Table 3). Calibration curves showed that both the RVOTD-VT/VF score and non-mortality score fit well with the corresponding risks at 1, 2, and 3 years (Supplementary Fig. 4). A consistent difference in the VT/VF rates was also observed between the primary and secondary prevention populations (Supplementary Table 4).

The RVOTD-ICD benefit score was then calculated and displayed a normal distribution (Shapiro–Wilk test, p = 0.109; Supplementary Fig. 5). According to the benefit score, this population was divided into the high benefit group ($>$61), intermediate benefit group (54–61), and lowest benefit group (54). Fig. 3A,B illustrates the cumulative incidence curves for the observed risk of VT/VF and non-arrhythmic mortality in each of the three RVOTD-ICD benefit groups. The benefit score divided groups well stratified the whole population in both the risk of VT/VF (Fine-Gray, p 0.001) and non-arrhythmic mortality (log rank, p 0.001). The ATP-included endpoints displayed similar results (Supplementary Fig. 6). The predicted risks in the RVOTD-ICD benefit groups are shown in Table 4. In the highest RVOTD-ICD benefit group, the 3-year VT/VF risk was approximately 8-fold higher than the corresponding risk of non-arrhythmic mortality (39.2% vs. 4.8%, p 0.001, respectively). In the intermediate group, the 3-year VT/VF risk was also higher than the risk of non-arrhythmic mortality; however, the difference diminished (29.4% vs. 10.9%, p 0.001). In contrast, the 3-year risk of VT/VF was similar to the risk of non-arrhythmic mortality (21.9% vs. 21.3%, p = 0.405) in the lowest RVOTD-ICD benefit group.

Fig. 3.

Incidence of VT/VF (A) and non-arrhythmic mortality (B) among the three RVOTD-ICD benefit score groups. ICD, implantable cardioverter defibrillator; RVOTD, right ventricular outflow tract diameter; VT, ventricular tachycardia; VF, ventricular fibrillation.

MADIT-ICD VT/VF, non-arrhythmic, and benefit scores were calculated as in the original research. The time-dependent receiver operating characteristic curve (Fig. 4) manifested that RVOTD-ICD VT/VF score yielded a more accurate prediction of both 1-year VT/VF (0.64 vs. 0.57, p = 0.029) and non-arrhythmic mortality (0.81 vs. 0.69, p = 0.006). NRI analysis also revealed that the RVOTD-ICD VT/VF score and non-arrhythmic mortality score could better reclassify 16.1% and 42.4% of patients for 1-year outcomes, respectively. The improvement was consistent in the two- and three-year observations shown in Supplementary Table 3.

Fig. 4.

Receiver Operating Characteristic Curve for the MADIT- and RVOTD-VT/VF score and non-arrhythmic mortality score. RVOTD, right ventricular outflow tract diameter; VT, ventricular tachycardia; VF, ventricular fibrillation; MADIT, Multicenter Automatic Defibrillator Implantation Trial.

Finally, decision curve analysis was applied to facilitate the comparison of long-term survival benefits between the scores (Supplementary Fig. 7). In the 5-year analysis, the RVOTD-ICD benefit score provided a larger net benefit across the range of all-cause mortality risk than the MADIT-ICD score. For example, at a threshold of 30% death risk, the RVOTD-ICD benefit score identified 4.4% additional all-cause mortality compared with the MADIT-ICD benefit score, without increasing the number of false positives (Supplementary Table 5).

The fact that only 7–30% of ICD recipients for primary prevention in clinical trials received appropriate shocks suggests the need for improved SCD risk stratification [16]. Among the many clinical variables proposed as potential predictors of SCD [17], LVEF is a nearly exclusive marker used in the clinical decision of ICD implantation because of its simplicity and reliability, which has been substantiated in several randomized control trials [18, 19, 20]. However, in many observational studies, most patients with SCD had normal or mildly reduced LVEF [5]. The sole use of LVEF may be underqualified for the identification of patients at risk of SCD in different cardiac conditions, especially in non-ischemic cardiomyopathy [20, 21].

Moreover, previous studies have focused on constructing risk assessment tools to facilitate risk stratification in primary prevention ICD recipients, while CHF patients with secondary implantation were rarely taken seriously, probably based on the myth that these patients would certainly benefit from ICD. However, in consideration of cost-effectiveness, especially in underdeveloped areas, those with secondary indications are the majority of ICD recipients in real-world settings. Still, prior studies have failed to identify a specific group that exhibits a higher or lower risk of recurrence of potentially life-threatening ventricular arrhythmias among this population [22, 23]. More importantly, another thought-provoking basis for studying patients with CHF both with and without prior sustained VT/VF is that other causes of mortality that are not arrhythmic-related or even noncardiac play an increasing role in HFpEF [24], which is unavoidably underrepresented in the primary prevention population. In addition, efficacious medications for heart failure with reduced ejection fraction have been less so at higher LVEF ranges, only decreasing the risk of HF hospitalization but not cardiovascular death in HFpEF. This reflects the burden of noncardiac comorbidities as LVEF increases and emphasizes the complicated cardiac and noncardiac mechanisms underpinning HFpEF. Therefore, weighing between VT/VF and non-arrhythmic mortality needs to be individually conducted to appropriately evaluate the potential benefit of ICD in patients with CHF across LVEF. Our study concluded that LVEF and left atrial diameter were more competitive in predicting non-arrhythmic mortality than life-threatening ventricular arrhythmia in patients with CHF. Consequently, this phenomenon calls for other specific arrhythmogenic predictors for precise ICD implantation.

Recent studies have attached importance to RV function as a determinant of arrhythmic outcomes, including SCD and ventricular tachyarrhythmia [11, 25, 26, 27, 28, 29]. RV dysfunction (usually assessed by RV ejection fraction) is independently predictive of sudden cardiac arrest or appropriate ICD therapy, even among those with an LVEF $>$35% [11, 26, 27], which indicates that RV dysfunction has the potential to enhance the current approach to SCD risk stratification beyond left heart function. The possible mechanism by which RV involvement leads to the occurrence of VA/SCD has not been elucidated. Macro reentry, mainly due to a conduction delay within the scar zone, is frequently noted in ischemic cardiomyopathy. RV stretch and volumetric load would prolong cardiac repolarization and refractoriness, which may create an unstable electrophysiological substrate, giving rise to an enhanced propensity for stretch-triggered or stretch-mediated ventricular arrhythmias [30].

Surprisingly, this study showed that RVOTD was superior to LVEDD for arrhythmia risk stratification. The response of RV to disease is a consequence of various combinations of volume overload and/or pressure, as well as intrinsic myocardial deficits, where the predominant abnormality may determine the clinical presentation and course. Because the vast majority of patients had left-sided heart failure and acute pulmonary arterial hypertension was excluded in this study, additional RV involvement may be regarded as a sensitive barometer of any “downstream pathology” affecting RV afterload due to abnormal pulmonary vasculature, or most likely, increased LV filling pressures [31], which appears to identify a group with a high risk of ventricular tachyarrhythmia.

Although RV systolic function is an important predictor of cardiovascular outcomes, the complex morphology of the RV and the mechanics of motion make it difficult to obtain accurate and reproducible measurements [32]. Since the assessment of RV function is not currently routine before ICD implantation, many echocardiographic measures related to RV function were not available for every patient in this study. Nonetheless, we proved that an integrated record of right ventricular diameter was useful enough to reflect RV stretch and loaded volume to a huge extent and undoubtedly had a significant implication on the prognosis of malignant arrhythmic events.

It should be noted that the long-term effect of the proarrhythmic action of RVOTD was consistent in most clinical subgroups, including a history of cardiovascular disease across LVEF strata, indicating that RVOTD could serve as an incremental predictor of SCD in patients with reduced systolic function, whose risk of SCD has already increased, as well as in patients with preserved LVEF. The exception lies in the use of three foundational drugs for the treatment of CHF: RAAS inhibitors, $\beta$-blockers, and mineralocorticoid receptor antagonists. Although patients with a higher RVOTD experienced more VT/VF events regardless of whether they were taking anti-HF medications, the former were clearly less influenced by a larger RV at the time of implantation. It could be speculated that these drugs improve cardiac remodeling in both ventricles, and thus, improve arrhythmic outcomes. Additionally, for those who cannot tolerate anti-HF medications, cardiac pump function is usually not able to maintain blood pressure. Consequently, capacity-sensitive RV tend to expand, which further increases the risk of long-term arrhythmic events.

Several baseline characteristics, such as atrial fibrillation and digoxin use, were imbalanced between the higher and lower RVOTD groups, which may have confounded the proarrhythmogenic effect of RVOTD. However, univariable and multivariable adjustments of these characteristics minimized the study bias. Moreover, during every interrogation, the preceding rhythm of each previous shock was examined through an intracavitary electrogram, thus preventing the mistake that inadequate shocks for atrial tachyarrhythmia were regarded as endpoints.

In light of the MADIT-ICD benefit score, the new RVOTD-ICD benefit score could be more easily calculated manually and used for decision-making regarding the need for SCD prevention. In this population, the benefit of ICD was obvious in the first two years among all candidates, while the efficacy started dividing in the third year. The highest RVOTD-ICD benefit group comprised patients with the highest predicted risk for VT/VF and the lowest predicted risk for non-arrhythmic mortality; hence, the absolute need to receive an ICD. In the intermediate benefit group, the 3-year risk of VT/VF was still nearly three times higher than the corresponding risk of non-arrhythmic mortality. Thus, they should also be considered for ICD prevention, combined with concomitant treatment of associated comorbidities, to reduce the risk of non-arrhythmic death. Although the risk of experiencing three-year VT/VF is still over 20% to warrant ICD implantation in the lowest benefit group, the significant difference from the risk of non-arrhythmic mortality vanishes, suggesting that a personalized approach to device implantation should be considered with more focus on the management of comorbidities to maximize the benefit from ICD. Furthermore, this score can be extrapolated to patients with previous ventricular tachycardia or fibrillation, who account for a large proportion of patients in real-world situations. Even though ICD implantation seems inevitable for these patients, the RVOTD-ICD benefit score could still serve as a reference for individualized follow-up, including the frequency and emphasis of interrogation, and for more urgent ICD treatment, prior to the progression of more advanced risk factors associated with non-arrhythmic mortality.

Apart from the RVOTD identified in this study, one of the most recognized risk factors associated with ICD efficacy is age, as younger patients have a higher probability of appropriate ICD therapy [14, 15, 23, 33]. Prior NSVT is another established hazard for VT/VF risk [34]. This component and prior sustained VT/VF were assigned different points in the RVOTD-ICD VT/VF score. A large prospective study supported the extension of ICD to a selected population with prior NSVT from underrepresented geographies [35]. Diuretics have not been shown to reduce all-cause mortality in patients with CHF. Instead, we indicated that the need for diuretics in patients with CHF implies worse prognosis. It is also worth noting that thiazide and loop diuretics may cause electrolyte abnormalities and concomitant drug-induced arrhythmias.

With regard to non-arrhythmic mortality, diabetes [36] and a higher NYHA class [37] are widely acknowledged to present a high risk for non-arrhythmic mortality associated with various comorbidities. A larger left ventricular dimension, featuring left dysfunction, showed a better determination of pump failure death owing to its direct pump effect [38]. RAAS inhibitors, as evidenced by many randomized clinical trials [39, 40], were the only type of medications that outperformed other candidates in score development, proving that their clinical application cannot be overemphasized. Higher NT-proBNP levels have been shown to be closely associated with pump failure death [41]. Our previous investigation also suggested that patients with higher NT-proBNP levels might derive less benefit from ICD [42]. In this study, log-transformed NT-proBNP showed a linear relationship with non-arrhythmic mortality; thus, quartiles of log-transformed values were given as cutoff points to appropriately quantify its hazard ratios.