Electrocardiographic risk stratification in patients with a primary prophylactic implantable cardioverter defibrillator: can future arrhythmias correlating with sudden cardiac death be predicted?

Graphical Abstract

Open in new tabDownload slide

This editorial refers to ‘Competing risks for monomorphic versus non-monomorphic ventricular arrhythmias in primary prevention implantable cardioverter defibrillator recipients: Global Electrical Heterogeneity and Clinical Outcomes (GEHCO) study’ by L.G. Tereshchenko et al.  https://doi.org/10.1093/europace/euae127 (published ahead of print 4 May 2024).

Prophylactic defibrillator treatment, clinical outcomes, and ventricular tachycardia ablation

Early landmark studies^1,2 have established the implantable cardioverter defibrillator (ICD) as routine treatment to improve the rates of survival in the primary prevention of sudden cardiac death (SCD). Two decades later, we realize that all-cause mortality and appropriate shock rates have decreased overall,^3,4 while the incidence of both can be quite variable with comorbidities.⁵ Patients with an ICD may never receive appropriate shocks because of a high risk of non-arrhythmic death or—in a different risk pattern with similar consequences—simply have a low risk for death and arrhythmias.⁶ Are all prophylactic ICDs necessary? The answer to this question is fuelling a long-term and ongoing debate. The DANISH ICD study⁴ was unable to demonstrate overall ICD survival benefits in patients with systolic heart failure and non-ischaemic cardiomyopathy (NICM). The PROFID EHRA trial⁷ funded by the EU will investigate clinical ICD effectiveness early after a myocardial infarction in the 2020s (with left ventricular ejection fraction, LVEF ≤35%). In addition to randomized studies for ICD indications, there is a need to devise concepts for patients at high risk of SCD and for those hidden in the large group of primary prophylactic ICD candidates. Currently, only a few risk stratifiers identify such patients and truly predict ventricular tachycardia (VT) or ventricular fibrillation (VF) but not mortality. Examples are programmed electrical stimulation⁸ or periodic repolarization dynamics from a 24 h Holter electrocardiogram (ECG).⁹ This research gap has to be reduced further. The current work by Tereshchenko et al.¹⁰ does this by extensively studying risk stratification of patients who may develop malignant ventricular arrhythmias after prophylactic ICD implantation to treat ischaemic cardiomyopathy (ICM) and NICM. The authors used digital 12-lead ECG markers of global electrical heterogeneity (spatial dispersion in total recovery time), also previously validated in patients.¹¹ The markers included the spatial QRS-T angle, the spatial ventricular gradient elevation, and the sum of the absolute QRS-T integral. A digital 12-lead ECG is a readily available tool, and the analysis is straightforward with computer algorithms storable on the internet.

The Global Electrical Heterogeneity and Clinical Outcomes study

The authors enrolled almost 2700 patients with ICM and NICM undergoing primary prophylactic ICD treatment for systolic heart failure. The retrospective study design enabled only six US academic centres to recruit this impressive number of patients. This was one prerequisite for compiling meaningful complex statistics on the data, and the other was that the enrolled cohort was representative in terms of patient characteristics. The patients were young (with a mean age of 63 years) and had a low LVEF (28%); there was a predominance for male patients typical for ICD cohorts (77%); there was a good proportion of patients with NICM (44%) vs. ICM; and the patients had single- or dual-chamber ICDs or cardiac resynchronization therapy-defibrillators. Virtual recruitment was done at the time of first ICD implantation. The patients were followed up for a mean period of 4 years, usually ending before generator replacement. Predefined clinical endpoints were sustained monomorphic VT (MVT), polymorphic VT or VF (PVT/VF), or death without prior antiarrhythmic therapies.

Global Electrical Heterogeneity and Clinical Outcomes study results and statistical analysis

The clinical endpoint most frequently reached was MVT in 359 patients (3.17% per year). Patients are usually treated more often with anti-tachycardia pacing (ATP) than with ICD shock. Polymorphic VT/VF was seen in only 129 patients (1.14% per year). However, all patients were treated with ICD shock. Another 54 patients had undefined sustained VT/VF (0.48% per year) as an intermediary between the above. Death without prior therapy was reported in 479 patients (4.40% per year). Unfortunately, additional deaths following ATPs or shocks were not reported.

In the statistical analysis, the endpoints MVT vs. PVT/VF did compete with each other, and the ECG risk stratifiers were strongly and independently associated with both endpoints [area under the curve (AUC) for MTV: 0.728, AUC for PVT/VF: 0.918]. The multivariate models included up to 35 variables, covering all known predictors of all-cause mortality,^5,6,12 parameters of ICD programming, timing of enrolment, and known ECG risk markers. Competing risk models, such as the Fine–Gray model, were compared with Cox regression. Calibration was satisfactory, but was clearly better for PVT/VF. Remarkably, at least a decile with an MVT or PVT/VF risk >50% could be predicted. These patients very likely belong to the high-risk group of patients with SCD who are sought for by clinicians for treatment. In the predictive models, MVT predictors were clearly different from those for PVT/VF, underscoring their possibly different myocardial arrhythmic substrates. Monomorphic VT is a typical arrhythmia for ICM. Validation of predictive models was done using 10-fold cross validation. Using this type of internal validation is appropriate. However, it remains unclear how the results generalize to other populations in terms of geography, time, and domain.¹³ Importantly, the statistical analysis corrected for ICD programming, which is a major factor governing the incidence of the endpoint arrhythmias. Furthermore, corrections were also done for enrolment in time, as cardiovascular outcomes improved in the course of the long timeline of enrolment in this study. The corrective statistical measures chosen by the authors were sophisticated; however, the fact that they did not fully correct for the effects of these powerful predictors cannot be overlooked. The corrections for time of enrolment accounting for era effects are interesting. As expected, the clinical endpoint death without prior ICD therapy decreased considerably in the course of time (from 5.98% per year during the first era to 2.57% per year after 2014). Interestingly, the competing risk prediction of MVT vs. PVT/VF was not influenced by this. For PVT/VF, the incidences remained the same over time, while the risk for MVT decreased over time, reflecting the improved overall treatment. This may be one explanation why recent large ICD cohort studies¹² confirmed the survival benefit in patients reported in the early landmark studies^1,2 despite lower mortalities overall.

Clinical consequences of the Global Electrical Heterogeneity and Clinical Outcomes electrocardiogram

It is a diagnostic and therapeutic innovation if patients can be identified as high risk for MVT and further treated with a prophylactic VT ablation. A main result of the Global Electrical Heterogeneity and Clinical Outcomes (GEHCO) study¹⁰ was that there may be an application of risk stratification for patients with high risks of sustained MVT. However, at this point, this is a study idea that must be further developed into a randomized clinical protocol with sufficient sample size and funding, to be eventually done and published. The Pan-Asia United States Prevention of Sudden Cardiac Death study¹⁴ by Tung et al. dealt with a similar issue and ascertained a positive clinical benefit for patients undergoing VT ablation prophylactically, but the results cannot be extrapolated. In this recent randomized study in 180 patients with ICD implantation, a prophylactic VT ablation before ICD implant reduced a combined endpoint of VT recurrence, hospitalization, or death by 42%. It, however, did not prevent VT recurrences as such in the treatment group; it only reduced them. For patients in the risk spectrum for PVT/VF and for the large number of patients who are eventually not treated by implanting an ICD, there is no such option of prophylactic treatment. However, the implementation of informed personalized decisions on these patients and the management of these patients can be based on the individual prediction score from this study.

The mortality benefit accruing from the ICD is transmitted through life-saving shocks when malignant ventricular arrhythmias occur. Treating sustained MVT by means of ATP may also be viewed as a prophylaxis of shockable arrhythmias. If prophylactic therapy against ventricular arrhythmias causing SCD is desired, a prophylactic ICD can be implanted. In addition, prophylactic VT ablation could potentially supplement ICD therapy if risk is identified as in GEHCO.¹⁰ Polymorphic VT or VF is not considered a possible option to ablate prophylactically.

Limitations of the study

The GEHCO study suffers from the limitations of a retrospective study. However, this study represented the best solution in terms of efficiency when initiating it. Furthermore, criticism may arise on the ground that all risk stratification rested on a single 12-lead ECG obtained before ICD implantation. The authors avoided very long observation periods of patients and concentrated on a post-enrolment period of 3 or 4 years. A real challenge posed by the data collection process was the change in outcomes in the course of time; typically, the rate of mortality without ICD therapy dropped from the first period to the last from 5.98 to 2.57% per year. This difference was accounted for in the statistical models. Apart from death, which is typically considered a competing event, it is not always straightforward to decide which event types are to be viewed as competing events in the analysis and which ones are to be considered censoring events, although such an exercise can substantially bias the risk estimation.¹⁵ With other events, such as device generator change/upgrade, which was dealt with as independent censoring by Tereshchenko et al., there is less clarity. Follow-up could have been continued following events such as device generator change/upgrade but this might have resulted in altered cause-specific hazards. Similar arguments can be made for MVT and PVT/VF that were considered competing events. Visualizing this type of complex data is not always straightforward, as Kaplan–Meier curves are not appropriate in the presence of competing events, but the cumulative incidence function (also known as Aalen–Johansen estimator) provides unbiased estimates of event probabilities in these settings.¹⁶ Finally, heterogeneity between centres was not explored, but it might provide some insights into the generalizability of the prediction models presented. Overall, we wish to congratulate Dr Tereshchenko and colleagues on a well-conducted study on a clinically very relevant topic.

Funding

This work was supported by the FP7 Health European Union FP7/2007-2013 (Grant agreement no. HEALTH F2-2009-602299) for 5 years (starting 1 October 2013).

Data availability

No new data were generated or analysed in support of this editorial.

References

Author notes