Implantation of internal defibrillators, capable of monitoring the heart and shocking life threatening arrhythmias back to normal rhythm, for patients with severe heart failure increases the probability of survival. Arrhythmias of the atria of the heart are common in these patients. Administering a direct current electrical shock under anesthesia (cardioversion) is the method of choice to reestablish normal sinus rhythm in this instance. Safety and efficacy of external electrical cardioversion (CV) in patients with ICDs was demonstrated in several studies. Safety of internal cardioversion (shocking the heart back into normal rhythm via the implanted defibrillator) was described in several smaller trials.

Performing external instead of internal cardioversion in patients with implanted ICDs is more feasible for most hospitals, as CV can be performed without a programming computer and an additional specialist present, e.g. on the intensive care ward, and device interrogation can be done after CV at the remote ICD/pacemaker clinic.

No scientific data on safety and efficacy endpoints comparing internal vs external CV is currently available. The aim of the study is to compare external vs internal electrical cardioversion for atrial arrhythmias and establish a safety and efficacy profile for external and internal cardioversion in large cohort of ICD patients.

Introduction A mortality benefit of implanted cardiac defibrillators in patients (pts) with severely impaired LV function has been established in multiple large trials. Atrial fibrillation, -flutter and -tachycardia are a common comorbidity in this patient population. Cardioversion is the method of choice to reestablish normal sinus rhythm in this instance. Safety and efficacy of external electrical cardioversion (CV) in patients with pacemakers and ICDs could be shown in several smaller single center trials. Safety of internal cardioversion was described in several smaller trials. Reports on efficacy of internal CV vary between 33% and 83% .

Performing external instead of internal cardioversion in pts with implanted ICDs is more feasible for most hospitals, as CV can be performed without a programmer present, e.g. on the ICU ward, and device interrogation can be done after CV at the remote ICD/pacemaker clinic. A working group of the German cardiac society recommends internal CV for patients carrying an ICD, suggesting a superior safety profile for internal CV, despite conflicting data and a lack of evidence and on this issue. Thus, posing a problem in logistics for many hospitals.

No scientific data on safety and efficacy endpoints comparing internal vs external CV is currently available.

Aim of the study To compare external vs internal electrical cardioversion for atrial arrhythmias and establish a safety and efficacy profile for external and internal cardioversion in large cohort of ICD patients. This study will provide vital evidence for future recommendations regarding cardioversion in the growing population of patients carrying implanted ICD and pacemakers. The main purpose will be to establish non-inferiority of external electrical cardioversion to internal electrical cardioversion. This would impact day to day practice in many arrhythmia centers and smaller hospitals around the world.

Trial populations and choice of primary analysis set The full-analysis set (FAS) is derived from the intention-to-treat (ITT) principle. This dataset includes all trial subjects randomized, with any valid data on the primary target variables "occurrence of severe lead and device impairment" and "restoration of normal sinus rhythm".

The per-protocol set includes all trial subjects who were essentially treated according to protocol (i.e. who have been cardioverted as assigned) and for whom the primary target variables "occurrence of severe lead and device impairment" and "restoration of normal sinus rhythm" could be determined based on complete and meaningful documentation.

In a non-inferiority trial there are special issues in the choice of primary analysis set. Conservatism or anticonversatism of the ITT or PP analysis depends on various factors as type of protocol deviation and missingness, and the method of handling missing data in ITT population (see Sanchez MM, Chen X. Choosing the analysis population in non-inferiority studies: per protocol or intent-to-treat. Stat Med. 25, 1169-81;2006). Due to our setting (randomization immediately before intervention, allowance for cross-over from internal to external CV, intervention in hospital setting) we do not expect a relevant amount of protocol violations. As the follow-up period of 2 weeks is short, we expect <1% losses to follow-up. So we assume only minor discrepancies between ITT and PP population.

We do not expect any missing data for the primary efficacy endpoint, as it is measured immediately after CV. The primary safety endpoint is measured within two weeks after CV, and we do not expect non-trivial missingness such as dropout due to lack of efficacy. As the expected number of events for the primary safety endpoint is low, especially for internal CV patients, any method of handling of missing data in ITT analysis may cause relevant bias. Thus, the per-protocol set is considered the primary analysis population; the ITT analysis is treated secondary/supportive and should lead to similar conclusions. Patients without any valid data on the primary target variables will be excluded from the FAS (modified ITT), in case of missing data on safety endpoint the valid value will be used. In a sensitivity analysis a conservative scenario will be performed treating missing values as event in external CV group and as no event in internal CV group. Any relevant discrepancies in drop-out rates between CV groups will be discussed.

Outcome data and data analysis In case of cross-over from internal to external CV the primary endpoints will be assessed after internal shock. Patient's death during intervention or within the follow-up period is unlikely and not to be expected. In the case of patient death during intervention, after the shock, the primary efficacy endpoint will be treated as failure. In the case of a patient death during intervention, after the shock or during follow up, the primary safety value will be counted as failure/severe device or lead impairment. This may be revised after the cause of death was proven not to be lead or device associated by interrogation and analysis by the manufacturer.

Stratification by study center is not feasible due to the low number of expected events, and subgroup analyses by study center will be done for descriptive reasons only.

Duration and study size This study is a multicenter trial. Additional local ethics application is required at each center.

For the primary safety endpoint "severe lead and device impairment" we expect an incidence of 10% in the external CV group and of 5% in the internal CV group. Thus, for one-sided alpha 5%, non-inferiority margin of 15%, and power 80%, a sample size of 103 patients per arm, i.e. 206 in total, is required. To account for loss to follow-up (<1% within 2 weeks after CV) and heterogeneity between centers, 230 patients (≈206/(1-0.05)^2) (see Donner, A. Approaches to sample size estimation in the design of clinical trials--a review. Stat Med 3, 199-214;1984) will be randomized. A non-inferiority margin of 15% is large, however, we chose several endpoints acting as sensitive surrogates of possible or temporary lead affectation, rather than relying solely on hard endpoints such as permanent lead or device failure with necessity of surgical revision.

As no data on placebo-controlled trials exists in the setting of ICD trials to determine the non-inferiority margin on statistical grounds, the margin was chosen on clinically as well as practicality grounds. A relevantly smaller alpha or tighter margin would lead to an unachievable sample size and render the trial unfeasible (see D'Agostino, R. B., Massaro, J. M. & Sullivan, L. M. Non-inferiority trials: design concepts and issues - the encounters of academic consultants in statistics. Stat Med 22, 169-186;2003).

Of note, to detect an absolute risk reduction of 20% in the primary efficacy endpoint "restoration of normal sinus rhythm", assuming an incidence of 90% in the external CV group and 70% in the internal CV group, a sample size of 72 patients per arm, i.e. 144 in total, is required (two-sided alpha 5%, power 80%, continuity corrected chi-squared test). Sample size determination was performed with ADDPLAN™ 6.0.

Description of trial subject groups At baseline (i.e. before CV) treatment groups will be described and compared using mean, standard deviation and percentiles (0, 25, 50, 75, 100) for continuous variables, count and percentage for categorical variables. Variables to be summarized include among other sex, age, BMI, comorbidities, type of device, site of implantation, LV ejection fraction, type of anticoagulation, duration of arrhythmia, left atrial diameter, pacing threshold, lead impedance, shock impedance, ventricular and biventricular pacing percentage, battery voltage.

Primary safety endpoint The primary hypothesis "non-inferiority of external CV to internal CV" regarding the primary safety endpoint "severe lead and device impairment" will be evaluated using the confidence interval (CI) approach. A one-sided 95% CI for the difference "observed proportion in external group - observed proportion in internal group" will be calculated by Newcombe's method 10 and non-inferiority is declared to be shown at one-sided level 5% if its upper bound lies below the non-inferiority bound of 15%.

Primary efficacy endpoint If non-inferiority regarding the primary safety endpoint "severe lead and device impairment" could be demonstrated, the hypothesis "superiority of external CV to internal CV" regarding the primary efficacy endpoint "restoration of normal sinus rhythm" will be tested in a confirmatory way at two-sided level 5%. Due to the fixed sequence of these two hypotheses, the type I error is not inflated. The proportion of the primary efficacy endpoint will be compared between treatment groups by Fisher's exact test with two-sided 95% CI calculated by Newcombe's method 10.

Secondary endpoints The components of the primary safety endpoint (i.e. a rise in threshold (at constant duration) of >0.5V, exit block, loss of programming, rise in shock impedance by 50%, rise in charge time by 50%, drop in battery voltage of ≥0.2V within 2 weeks) will be analyzed along the same line as the composite.

Moreover, recurrence of atrial arrhythmia at FU, lead impedance and sensing changes, inadvertent induction of VF during CV, high-sensitive Troponin T level 3h after CV will be summarized by treatment group and compared by statistical hypothesis tests as applicable; confidence intervals for treatment differences/effect sizes will be calculated to aid interpretation.

Subgroup analysis

A subanalysis of predefined subgroups will be performed for:

• CRT-D patients to assess safety of external CV in pts with LV leads.
• CRT-D patients with quadripolar leads (including thresholds for all 16 pacing vectors)
• CRT-D patients with AV-VV optimization algorithms

Safety and suspension/termination criteria Based on published data and own previous work 10 an interim safety analysis will be performed after 50% of target sample size, i.e. 115 pts, have been included. A statistically significant difference in the primary safety endpoint between the groups at interim analysis will lead to a subsequent early termination of the trial (using Fisher's exact test, two-sided level 5%).

Specific termination criteria:

• Significant difference in primary safety endpoint at interim analysis
• Induction of VF in internal CV group n > 3

Inclusion Criteria:

• Age ≥ 18 years
• Informed, written consent
• Atrial arrhythmia with indication for CV
• Status post ICD implantation, including CRT-D

Exclusion Criteria:

• Age < 18 years
• Patients under guardianship or with mental disorders / disabilities
• ICD implantation < 4 weeks prior to CV
• ICD lead implantation < 4 weeks prior to CV
• Battery in EOL, ERM or ERI, ERT
• Indications of compromised leads (Impedance <200 or >2000 Ohm, Pacing threshold >5V/0.4ms), RV Sensing <4mV or RA sensing <0,1mV)