Statistical analysis was performed with R (version 4.0.2) and SPSS™ (version 23). Continuous variables are expressed as mean values (± one standard deviation). Categorical variables are expressed as numbers (percentage). T-test and χ² test were used where appropriate. Calculation of the cumulative event-free survival was performed with the Kaplan-Meier survival function and the Log Rank test was used to examine difference between groups. Annualized event-rate was calculated by dividing the number of events by patients-years (patients at risk multiplied by mean follow-up). No multivariable logistic regression model to explore for predictors of arrhythmic events could be performed since the event number was too low.

In the four participating centres, 66 patients with super-response were identified. Baseline characteristics of the patients are displayed in Table 1. First implantation of CRT-D was between 02/2000 and 04/2018. In downgraded patients, diabetes mellitus was more prevalent (33% vs. 10%, p = 0.05) and the QRS duration measured on 12-lead electrocardiogram (ECG) was significantly shorter (158 ms vs. 172 ms, p = 0.01). There were no other statistically significant differences between the groups.

Median overall follow-up after first CRT-D implantation was 129 months (IQR: 101–155). Median follow-up after downgrade or first GE was 53 (IQR: 30–82) months. First generator exchange was performed after median 64 (IQR: 51–71) months. 21 patients (32%) underwent a second GE after median 57 (IQR: 48–70) months. At first GE, mean baseline LVEF had improved from 25% ± 7% to 53% ± 6% (p < 0.001). 57 (86%) patients experienced super-response at first GE. 9 patients (14%) did not achieve super-response until the second GE, although it should be noted that mean LVEF in these patients was considerably higher compared to baseline at the time of first GE (43% ± 3% vs. 26% ± 5%, p < 0.001). In 12 patients (18%) LVEF was assessed at the last GE. In patients with available LVEF recently before file closure, 85% remained super-responders (46/54 patients).

During the follow-up period, 27 patients (41%) were downgraded from CRT-D to CRT-P. The other 39 patients (59%) had replacement CRT-D generators implanted at GE. The median time to downgrade to CRT-P was 68 months (IQR: 58–98) after implant. Mean LVEF at the time of downgrade was 54% ± 4% – this had improved significantly from baseline (24% ± 7%, p < 0.001). 21 patients (78% of those downgraded) were downgraded to CRT-P at first GE, whilst the other 6 patients (22%) were downgraded at second GE.

Two arrhythmia events (3%) occurred in the first 12 months following initial CRT-D implant (censor period). This period following implantation is assumed to be the time of greatest myocardial remodelling (9). •

Monomorphic VT after one month in a female patient (60 y) with non-ischemic cardiomyopathy (NICM); LVEF 26% at baseline; terminated by ATP; treatment with amiodarone was started until 8 months (stopped for intolerance);

Out of those downgraded to CRT-P, no significant arrhythmia event occurred during a median follow-up of 38 (IQR: 29–53) months after DG as assessed by device interrogation. However, three events (8%, 3/39) were observed in the patients who had CRT-D replacement after the first GE over a median follow-up of 70 (IQR: 39–97) months (Table 2). Survival free of significant arrhythmia is depicted in Figure 1. The difference in event-free survival was not statistically significant (Log Rank p = 0.195). There was no difference in overall survival as shown in Figure 2 (Log Rank p = 0.868).

Annualised event-rates after DG/GE amounted to 1.5%/year in the CRT-D group, reducing to 1.0%/year when considering the whole cohort. The annualised event-rates during total follow-up were 0.7%/year in the CRT-D group and 0.4%/year in the whole cohort.

As shown in Table 3, 8 (12%) deaths occurred after median 90 (IQR: 80–108) months, only two (25%) of these were cardiac. No cardiac death occurred in downgraded patients.

Ogano et al. demonstrated the feasibility and mid-term safety of downgrading responders (defined as LVEF ≥45%, primary prevention, no VT/VF since implantation) (14). From this cohort of 49 consecutive patients, 7 (14%) were downgraded at GE and followed-up during an observation period of 40 ± 21 months. Similar to our cohort, downgraded patients experienced no arrhythmic events, although it should be noted that the number of downgraded patients in this study was relatively low. Additionally, the downgraded patients tended to have a lower all-cause mortality compared with those patients not deemed to be responders. This trend was not present in our cohort, which may be because all patients assessed in this study fulfilled the criteria for super-response.

Garcia and colleagues downgraded 14 patients to CRT-P following improvement of LVEF to >35% with no ICD therapies (13). Mean LVEF at GE was 49%. During a follow-up of 5.1 ± 1.3 years, there were 2 incidences of VT, but no sudden death. The research team concluded that downgrade was a safe and cost-effective treatment option. Comparability to our cohort is limited as the definitions of super-response differ significantly and the follow-up is twice as long.

Event rates in our study were surprisingly low for such a long follow-up period, although low event rates in super-responders were also seen in the study by House et al. (15). In 30 ICD patients with super-response (LVEF ≥ 50%), no arrhythmias were seen during a mean follow-up of 25 ± 18 months after generator exchange. However, this study has several key differences compared with our study: the follow-up period was shorter and the sample size smaller, and the cohort included 16 CRT-D patients only.

Nesti et al. observed 103 CRT-D patients for 26 ± 10 months after GE (16). Responders were defined as decrease in left ventricular end-systolic volume of ≥15%. Four responders (4% of the cohort) experienced a first arrhythmia after GE. Unfortunately, it is not reported how many of these responders who suffered from arrhythmia events were super-responders with LVEF ≥50%, which would allow direct comparison with our cohort.

Arrhythmic events are rare in super-responders as shown by Ghani et al.—in this cohort of CRT-D patients followed until their first GE, none of those classified as super-responders (56/347, 16%) experienced an ICD therapy over a follow-up of 5.3 years (4).

Zecchini et al. showed that in the super-responder subgroup (24%, 62/259), only 7% had a first appropriate device therapy during a mean follow-up of 68 ± 30 months despite LVEF >50%, resulting in an annualized event rate of 1.2%/year (6). Using a similar definition of super-response, Killu and colleagues demonstrated a lower event rate of ICD therapy (0.4%/year, and a cumulative 5 year rate of 2.7%) (5). A recent meta-analysis revealed a rate of ventricular arrhythmia of 0.9%/year (17). Thus, all studies are in line with our findings (annualized event rate of 1.0%–1.5%/year after DG/GE, 0.4%–0.7%/year during total follow-up). With our study looking at a large cohort of downgraded super-responder patients over a longer follow-up period, this study adds to the growing body of evidence that downgrading to CRT-P appears safe and could be considered by clinicians.

Certain patient characteristics are associated with super-response, such as female gender, NICM, higher LVEF at baseline, LBBB morphology, wider QRS duration, BMI <30 kg/m² and smaller baseline atrial size (4, 18).

Primary prevention defibrillator was not associated with a significant lower mortality compared to a combination of CRT-P and medical therapy in symptomatic patients with NICM (DANISH-trial) (12). The majority of patients (74%) in our cohort had NICM and therefore might represent patients similar to the DANISH-trial at their first GE. As ICD was superior in the subgroups of age <59 years and NT-proBNP <1,177 pg/ml only (12), these factors should influence choice of device at GE. Furthermore, survival advantage with ICD is less clear in older (≥75 years) and diabetic patients (EU-CERT-ICD) (19). In addition, extensive myocardial scaring on cardiac magnetic resonance is a known risk factor for the occurrence of arrhythmic events (HR 5.2) (20).

Although these predictors were established at the time of initial CRT-D implantation, knowledge of these is important to guide and counsel patients at the time of GE. A clinical score usable at GE incorporating baseline and follow-up factors (e.g., age, gender, frailty, malignant arrythmia, myocardial scar, comorbidities) would be desirable. Our sample size and event rate was too small for solid assessment of such factors. But further sub-studies from larger cohorts, meta-analyses or eventually data from the RESET-CRT trial could be useful (21).

Potential advantages of a downgrade are no inappropriate ICD-shocks [which can be up to 20% (11)], lower infection rates (12) and smaller pocket size. Furthermore, the longer battery life, wider control interval and lower costs can be beneficial for certain healthcare systems. These factors gain more importance in patients with super-response and thus low risk for arrhythmic events. In a contemporary cohort study with CRT-D systems implanted after 2015, the percentage of inappropriate therapy and inappropriate shock were still 7.2% and 3.3%, respectively (22).

Different CRT-D header types exist which complicate the option of downgrading. In systems using DF-1 headers, high-voltage parts can simply be disconnected, and the pace-sense part be connected to a new CRT-P device. Despite being widely replaced by DF-4 leads, still a considerable number of patients have a functioning DF-1 lead and present themselves for GE. In patient with newer DF-4 leads, downgrade to CRT-P is not possible without implanting a new pace/sense lead and abandonment of the DF-4-ICD lead. This additional lead implantation could result in potential complications which might outweigh the benefits of a downgrade. Another possible, but manageable downside of a downgrade is that the CRT-system loses its MRI compatibility if the old ICD lead is not extracted.

Modern pharmacological treatment markedly improved mortality and morbidity of patients suffering from heart failure (23). With the addition of two potent agents (angiotensin receptor/neprilysin inhibitors (ARNIs) and sodium–glucose co-transporter 2 (SGLT-2)) many patients nowadays have the chance to profit from these “fantastic four”. All these agents improve cardiovascular outcomes, and even reduce arrhythmia and sudden cardiac death (24–27). Furthermore, accumulating data question the mortality benefit of ICDs in primary prevention, especially in patients with NICM (12, 19, 28). Hence, it can be anticipated that the clinical scenario with super-responders presenting for device replacement will occur more often, and consequently the question regarding downgrading to CRT-P. Additionally, with rising health care costs, also economical pressure might influence this decision in the future.

Due to the retrospective design of this study, limitations associated with this study design are possible. Due to the lack of randomisation and a predefined protocol, this study represents real-world follow-up data with no predefined timepoints for device interrogation and echocardiography (no core lab available). But the study is based on clinical data which is also used for decision making in clinical routine.

Since not all study sites have a prospective registry, percentage of super-responders from the overall cohort could not be determined. Therefore, selection bias cannot be excluded.

The decision to downgrade was taken in consent between patient and treating physician and not in a randomised fashion. There is no data available for pharmacological heart failure treatment regarding intensity, duration and agents used. This limits the further analysis as optimal medical therapy itself was also shown to reduce mortality and prevent arrhythmia. Nevertheless, all patients were treated according to the current guidelines which includes optimal medical treatment.

To optimally analyse the debate about downgrade of super-responders, a randomized-controlled trial would be optimal. But such a study is very unlikely to ever be conducted. Hence despite the above-mentioned limitations, our study provides important long-term prognostic data on a special subgroup of CRT-patients. Since the majority of these patients had their device implanted before the novel DF-4 header was introduced, still a considerable number of these super-responder patients present with an active DF-1 lead. Knowledge of their long-term course can help clinicians in their daily decision making.

Although reverse remodelling with super-response seems to be persistent over time (29), there are still patients who can deteriorate during follow-up. In our cohort, 85% of patients remained super-responders.