OUP user menu

Regulatory pathways for development of antiarrhythmic drugs for management of atrial fibrillation/flutter

Irina Savelieva, Paulus Kirchhof, Nicholas Danchin, Pieter A. de Graeff, A. John Camm
DOI: http://dx.doi.org/10.1093/europace/eur181 1063-1076 First published online: 2 August 2011


New antiarrhythmic drugs are urgently required for the treatment of atrial fibrillation (AF), an increasingly common sustained cardiac arrhythmia seen predominantly in the elderly population. Pharmaceutical companies are generally interested in this important group of patients and a relatively large number of antiarrhythmic agents are under development for several indications relating to AF, predominantly rhythm and rate management. Because of the significant clinical consequences of the arrhythmia, it has been recognized that controlled trials in patients with AF should assess the effect of therapy in several major outcome domains such as mortality, morbidity, and hospitalization, with an emphasis on stroke and heart failure. As part of a regular series of meetings, the European Society of Cardiology recently met with European regulators and representatives of the pharmaceutical industry to review the current status of medical therapies for AF. Special attention has been paid to the debate on the relevant clinical endpoints in future AF trials and their implications for drug indications. The need for large-scale major cardiovascular outcome and comparator studies for the approval of drugs designed to manage rhythm and/or control the rate has been discussed. The requirements for appropriate ancillary studies, including quality of life and left ventricular function assessment and cost-effectiveness analysis, have been identified. This article reports the discussions that were held.

  • Antiarrhythmic drugs
  • Atrial fibrillation
  • Atrial fibrillation burden
  • Clinical trial
  • Drug approval
  • Drug development
  • EMA
  • Hospitalization
  • MACE
  • Mortality
  • Outcome
  • Rate control
  • Rhythm control

Drug development in atrial fibrillation and flutter: scope of the problem

Atrial fibrillation (AF) is an increasingly prevalent condition within the cardiovascular arena, which now stands at epidemic proportion posing a significant burden on healthcare resources. Lifetime risk for the development of AF is 1 in 4 for men and women 40 years of age and older.1 From the epidemiological data on AF, it has been estimated that more than 6 million people are currently affected by the arrhythmia in Europe, and the projected number of patients with AF is set at least to double in the next 40 years.24 AF is not an innocent bystander of the old age and is associated with significant and costly morbidity, such as stroke and heart failure.

Treatment strategies for AF have traditionally been split into rhythm control and rate control.5 There is also an important element related to stroke prevention. In addition to this, therapies are directed at control of symptoms, improvement of quality of life, and prevention or reduction of major cardiovascular outcomes including death and stroke. Currently, rhythm control is not part of the most usual therapeutic regimen in general medicine or general cardiology, particularly in elderly patients, but some cardiologists and most cardiac electrophysiologists hold a different view despite the results of randomized clinical trials. Rate control as a primary strategy is now accepted in older, sedentary, and asymptomatic or mildly symptomatic [i.e. those with European Heart Rhythm Association (EHRA) class I−II] who have had their arrhythmia for many years, without significant impairment of ventricular function and exercise tolerance.6 Rhythm control remains an important first choice in patients who are highly symptomatic, in patients with recent onset AF, and in young and active individuals. Furthermore, patients seeking symptom relief in addition to optimal rate control often require rhythm control interventions.

Success of pulmonary vein isolation or left atrial ablation in restoring and seemingly maintaining sinus rhythm has prompted renewed interest and further impetus to the concept that rhythm control is beneficial if achieved effectively and safely. At present, for many patients, non-pharmacological therapies form part of a hybrid management plan where antiarrhythmic drugs are crucial in maintaining the effects that have been produced by ablation, not in the least because of a late attrition rate of 5–15% per year even after initially successful ablation.79 Thus, a rhythm control strategy will become more popular if antiarrhythmic drugs with good adverse effect profiles and a neutral or beneficial effect on survival and major cardiovascular outcomes become available. However, any new antiarrhythmic drug will need to demonstrate its clinical value in the context of evolving therapeutic options (ablation, improved antithrombotic therapy, ‘upstream therapy’).

In 1996, the European Medicines Agency (EMEA) published a document on regulatory guidance for the development of antiarrhythmic drugs in general.10 However, since then, the development of antiarrhythmic drugs for ventricular tachyarrhythmias and some forms of supraventricular tachycardia (e.g. atrioventricular nodal reentry tachycardia, atrioventricular reentry tachycardia, and isolated typical atrial flutter) has been effectively halted because highly effective non-pharmacological therapies such as implantable cardioverter-defibrillators and curative ablation-based techniques have become available for these arrhythmias. Thus, AF is the prevailing target for antiarrhythmic drug development. Consequently, a raft of new agents for the treatment of AF is under development, including amiodarone analogues and other multiple channel blockers. In addition, there are many new compounds with fundamentally novel mechanisms of action, some of which are in clinical phase II development.11 Furthermore, because ventricular rate control is pertinent to all forms of AF including those managed with rhythm control, new agents that address purely rate control or antiarrhythmic drugs with rate-controlling properties are under investigation. There has been no guidance that would deal with this aspect of drug therapy for AF. Therefore, guidance, which specifically addresses the regulatory issues of drug development for AF is needed.12

Selection of relevant outcome parameters in antiarrhythmic drug trials

The main outcome parameters that are relevant for the regulatory authorities include clinical, biochemical, and other indicators of ongoing biological and pathogenetic processes (usually called biomarkers), indicators of pharmacological response to therapeutic interventions, which could be in certain circumstances substituted by surrogate outcomes, which are usually biomarkers that relate to clinical endpoints. Clinical endpoints can be a clinical characteristic or a clinical variable that reflects patient well-being and ultimately, survival.

Older AF trials focused on a ‘surrogate’ endpoint of electrocardiogram (ECG) documented AF recurrence. However, AF may have significant consequences in affected patients, such as increased risk of death and stroke, and higher rates of acute coronary syndrome and heart failure. As evidenced by major outcome results in several large randomized trials in AF, the annual rates of death, stroke, myocardial infarction, congestive heart failure, and hospitalizations are ∼4.0, 1.5, 1.0, 2.0, and 20% per annum, respectively, despite optimal medical therapy which is usually ensured within the clinical trial.5 However, endpoints such as time to first recurrence ignore the possible cardiovascular outcomes benefit of continued treatment (Figure 1). Given the multiple clinical consequences of AF, from stroke to mortality, it is therefore recommended that the effect of new antiarrhythmic drugs on these outcomes should be tested. The development of a specific clinical ‘model’ in which it would be ethical to try an investigational agent for AF, similarly to the post-myocardial infarction or heart failure ‘models’ developed for studying drugs for ventricular arrhythmias, is essential.

Figure 1

Relevant endpoints in the antiarrhythmic drug trials based on the natural course of atrial fibrillation. Note that ‘surrogate’ endpoints such as time to first recurrence ignore the possible cardiovascular outcome benefits of continued treatment.

New proposals have emphasized the problem of complex and often poorly established associations between the antiarrhythmic effect (i.e. maintaining sinus rhythm) as a surrogate endpoint and patients’ symptoms, quality of life, exercise tolerance, and general health.13 Symptoms and quality of life are regarded by new proposals as secondary outcome parameters in the application for a marketing authorization and cannot be introduced as an indication.

Progression of atrial fibrillation

The concept of AF as a progressive disease is based on the evidence of continuous remodelling of the atria caused by AF itself, changes associated with ageing, progression of underlying heart disease, and genetic and environmental factors. Atrial remodelling occurs in both ‘lone’ AF and AF associated with cardiovascular disease, but it is more pronounced in the latter. Progression of AF can be reflected in the change from first diagnosed or recurrent paroxysmal AF to persistent or permanent AF which occurs on average at the rate of 5–15% per year,14,15 depending on a number of factors such as the presence of underlying heart disease, e.g. heart failure, hypertension, and obstructive pulmonary disease (Table 1). Patients with ‘lone’ AF show the lowest progression rates of ∼1–2% per year.16 Prevention of progression of AF to a more sustained variety is a clinically relevant outcome, and has been employed (mainly as a secondary endpoint) in some antiarrhythmic drug trials.17

View this table:
Table 1

Rates of progression of paroxysmal atrial fibrillation to persistent or permanent atrial fibrillation

StudyNumber of patientsAge, yearsType of AFFollow-up, yearsProgression of AF, %Predictors of progression (risk)
European Heart Survey, 2010121964 ± 13Paroxysmal; lone AF: 17%115 Permanent AF: 8 In subgroup with lone AF: 7 (persistent or permanent)Age >75 years (1.57), heart failure (2.22), hypertension (1.52), stroke/TIA (2.02), COPD (1.51)
Sakamoto, 1995137No progression: 62.4±11With progression: 70.1±8.2First detected paroxysmal1Sustained AF ≥6 months: 22Age ≥65 years, heart failure, CTR ≥50%, diabetes, LA ≥38 mm, LVEF ≤0.76, f waves in V1 ≥2 mm
UK GPRD, 2005418Men: 67 ± 11, women: 73 ± 10First detected paroxysmal; no co-morbidity: 32%2.711 at 1 year 17 at 2.7 yearsValvular heart disease (2.7), moderate-to-high alcohol intake (3.0)
Al-Khatib, 201023160 ± 13Paroxysmal; lone AF: 41.6%48 at 1 year 18 at 4 yearsAge (1.82 per decade), AF at presentation (3.56)
Pappone, 200810657.5 ± 11.5First detected paroxysmal; lone AF: 51% 5Recurrent paroxysmal: 52.8 Persistent: 53.3a Permanent: 35.5a In subgroup with lone AF: 3.7 (persistent), 1.8 (permanent)Age (1.19), heart failure (11.2), diabetes (17.3), drug therapy vs. ablation
CARAF, 200575764 (median)First detected paroxysmal88.6 at 1 year 24.7 at 5 years Any recurrent AF: 63.2 at 5 yearsAge (1.4 per decade), cardiomyopathy (2.41), aortic stenosis (3.04), mitral regurgitation (1.69), LA enlargement (3.05–4.17)
Danish Study, 198642666 (median)Paroxysmal9 (median)33.1Underlying heart disease, thromboembolism
Kato, 200417158.3 ± 11.8First detected, paroxysmal1457 at 10 years 77 at 15 yearsAge (1.27 per decade), myocardial infarction (2.33), valvular heart disease (2.29), LA enlargement (1.39)
Olmsted County, 20077144.2 ± 11.7Lone AF: 48% paroxysmal, 52% persistent25.231 (30-year probability: 29)bAge (1.7 per decade), QRS abnormalities (3.2)c
  • AF, atrial fibrillation; CARAF, CAnadian Registry of Atrial Fibrillation; COPD, chronic obstructive pulmonary disease; CTR, cardiothoracic ratio; GPRD, general practice research database; LA, left atrium; LVEF, left ventricular ejection fraction; TIA, transient ischaemic attack.

  • aIn patients on antiarrhythmic drugs (n = 45); bin the majority of patients within 15 years; cQRS ≥ 110 ms, QRS notching, small R in the precordial leads.

Different cardiovascular pathologies produce specific substrates for AF which may follow different evolutions and are likely to require mechanism-specific therapeutic strategies. However, antiarrhythmic drugs have never been systematically tested with regard to how efficacy might depend on the nature of any underlying cardiovascular disease. It is usual to tailor the antiarrhythmic drug to the individual patient based on specific underlying heart disease and the likely mechanisms of AF, but this is undertaken only to optimize safety. Of note, newer imaging techniques [e.g. magnetic resonance imaging (MRI)]18 may in the future help to define populations which should or should not be treated by a specific line of drug therapy. Scar imaging in the left atrium or left atrial function may in the future also serve as surrogates to assess treatment success and could serve as an endpoint for drug therapy, e.g. visualization and quantification of the amount of fibrosis as a reflection of atrial remodelling may be used to assess the effect of drug therapy on abolition of reverse of atrial remodelling.

Should atrial fibrillation and atrial flutter be distinguished?

Although typical atrial flutter has an anatomically well-defined macro-reentrant circuit in the right atrium as its major mechanism and therefore can be relatively easily cured by ablation, AF and atrial flutter are closely related.1921 The incidence of isolated atrial flutter is relatively low (<1%).22 AF and atrial flutter usually coexist and patients with flutter develop AF even subsequently to successful ablation (23.1–52.7%).21 The different mechanisms of arrhythmia perpetuation suggests that there may be a different response to antiarrhythmic drugs with regard to cardioversion to sinus rhythm, and in these studies atrial flutter and AF should be distinguished.12 Conversely, AF and flutter probably respond to treatment aimed at prevention of recurrence or improvement of survival and reduction of cardiovascular complications in a similar way, and can be treated as one entity in trials designed to investigate these outcomes.

Major cardiovascular outcomes

Mortality and morbidity

Outcome studies in AF are different from outcome studies in other clinical settings (e.g. hypertension), probably because of greater complexity of action of antiarrhythmic agents compared with antihypertensive or lipid-lowering drugs and probably because AF patients may be at higher risk than some other populations. In the past, the regulatory authorities have not needed to consider major outcome studies in an AF population because antiarrhythmic drugs have been generally investigated for both atrial and ventricular indications. Studies in patients at risk for sudden death have been sufficient to document potential safety issues. However, a large-scale major cardiovascular outcome trial in a high-risk AF population must now be considered prior to the approval of an antiarrhythmic drug solely for AF.12 All-cause mortality is currently recommended as part of any primary endpoint in a large outcome studies. Cardiovascular mortality is an essential composite of total mortality, although it may sometimes be difficult to define. Independent adjudication is essential for the distinction.

The ATHENA (A placebo-controlled, double-blind Trial to assess the efficacy of dronedarone for the prevention of cardiovascular Hospitalization or dEath from any cause in patieNts with Atrial fibrillation and flutter) trial is the only current example of a trial of this type which has demonstrated reduced cardiovascular mortality with dronedarone even in patients who remained in AF throughout the study.23 Assessment of mortality/morbidity benefits in antiarrhythmic drug trials should be an endpoint independent of any antiarrhythmic efficacy endpoint. This does not imply that a typical efficacy study should be powered to detect a small difference in mortality, but data on mortality need to be collected on an ‘intention-to-treat’ basis from the time of randomization. Of course, selecting an overly high-risk trial population may lead to adverse results (e.g. the hazardous effect of dronedarone in the heart failure population).24 Atrial fibrillation-related mortality is a specific and, therefore, highly relevant outcome, but the exact definition and means of proving it are lacking in AF patients.

Morbidity outcomes should include disease-related morbidity, such as new onset or worsening heart failure, stroke, and myocardial infarction. Ischaemic and haemorrhagic strokes should be recorded separately and preferably be confirmed using brain imaging methods (MRI or computed tomography); the severity of stroke and neurological outcome and the presence and intensity of anticoagulation need to be documented.12 Morbidity endpoints are essential if there has been no increase in mortality or no fatalities have occurred. Inclusion of high-risk patients (e.g. with congestive heart failure and left ventricular dysfunction, ventricular arrhythmias) is required as this would provide important safety data. All events included in the primary outcome should be fully adjudicated.


It is appropriate to include an estimate of hospitalization, as part of a composite outcome combining mortality (e.g. all-cause mortality), and/or other non-fatal cardiovascular events. Time to hospital admission, duration of hospitalization, or the number in-hospital days may not be suitable as an isolated primary outcome because hierarchically higher events such as death may reduce the number of hospitalizations or shorten the duration of hospitalization, but need to be recorded. The reason for hospitalization (e.g. cardiovascular) and the definition of cardiovascular hospitalization should be clearly defined because some types of cardiovascular hospitalization may be more closely related to subsequent death.25,26 Analysis of the AFFIRM (Atrial Fibrillation Follow-up Investigation of Rhythm Management) trial has shown that cardiovascular hospitalization was highly predictive of death, and no interaction between this endpoint and treatment assignment was observed.26 The association between cardiovascular hospitalization and subsequent death has also been reported in the Stockholm Cohort Study of Atrial Fibrillation, which showed that individuals who spent >2% of time in hospital with a cardiovascular diagnosis had significantly higher mortality than those who had spent less time in hospital (36.0 vs. 8.2 deaths per 100 patient years).27 Time in-hospital >2% increased risk of subsequent death by 2.5-fold, and re-hospitalization for cardiovascular reason within 3 months increased risk of death by 1.4-fold in the early cohort and by 2.43-fold in the later cohort for which a better account of therapy was available.

In AF populations, it is important to differentiate hospital admissions for treatment of AF (e.g. cardioversion) from hospitalizations for cardiovascular adverse events (e.g. proarrhythmia, progression of heart failure, stroke, or myocardial infarction) and from hospitalizations for non-cardiac causes. Furthermore, a reduction in AF-related hospitalization may be due to an antiarrhythmic effect (i.e. maintenance of sinus rhythm) or due to improvement in patient's symptoms (e.g. when the drug renders previously symptomatic AF asymptomatic). In other words, reduction in AF-related hospitalizations can be a consequence of an antiarrhythmic efficacy or a surrogate for improved morbidity. In this respect, the issue of separation of the ‘antiarrhythmic’ indication from the ‘outcome’ indication needs to be considered. A reduction in hospitalizations can, therefore, be viewed as a measurement that integrates efficacy and safety.

Time to first recurrence of atrial fibrillation/flutter

Time to first recurrence of AF or atrial flutter as a ‘surrogate’ outcome may provide direct evidence of the antiarrhythmic efficacy of the drug and has been the primary outcome of choice in many trials of antiarrhythmic drugs (Table 2). The majority of high-quality studies of antiarrhythmic drugs have been conducted in patients with persistent AF because the recurrence of a persistent episode is likely to occur during the first year and because of its sustained nature (as opposed to a brief paroxysm) it is easier to recognize and document. An advantage of the time to first recurrence endpoint is the relative simplicity of the study protocol and assessment of the effect. A recurrence has usually been defined in terms of AF of a certain duration, e.g. at least 30 s or >10 min or, in the case of paroxysmal AF, progression to persistent or permanent AF (Figure 1). Time to first recurrence of any AF and time to first symptomatic AF recurrence should both be analysed. The occurrence of persistent AF may have more significant implications for the patient's management (e.g. cardioversion, resumption of anticoagulation prior to cardioversion, and possibly, the change of an antiarrhythmic drug). Thus, progression to persistent AF per se may represent a clinically relevant endpoint in drug trials.17,28,29

View this table:
Table 2

Randomized controlled studies of pharmacological prevention of atrial fibrillation

StudyNumber of patientsDrugDesignType of AFRhythm at inclusion and cardioversionPrimary endpointFollow-upAF monitoring
UK Propafenone, 199548Propafenone (2 doses)Double-blind, placebo-controlled, dose-escalating, cross-overParoxysmal, AF ≥2 episodes on TTM during the observation period (3 months)Sinus rhythmTime to first symptomatic and ECG confirmed recurrence of AF or an adverse eventObservation period (3 months), low-dose and high-dose phases (3 months each, with 1-week half-dose period)ECG during symptoms
Benditt, 1999253Sotalol (3 doses)Double-blind, placebo-controlled, dose-ranging, parallel-groupParoxysmal or persistent; documented in the previous 3 monthsSinus rhythmTime to first symptomatic recurrence of AF or flutter after achieving steady state; secondary endpoints: time to first recurrence after the first dose, proportion of patients without recurrence at 6 and 12 monthsMaximum 12 monthsECG at 1, 3, 6, 9, and 12 months; TTM every 2 weeks and during symptoms
SOCESP, 1999121Sotalol vs. quinidineOpen label, no placeboPersistent AF <6 months stratified by AF ≤ 72 h and >72 hSinus rhythm for 24 h after pharmacological conversion or DCCMaintenance of sinus rhythm; secondary endpoint: characteristics of AF recurrenceMaximum 6 monthsECG and 24 h Holter monitoring at 1 week, 1, 3, 6 months; TTM every 2 weeks and during symptoms
CTAF, 2000403Amiodarone vs. sotalol or propafenoneOpen label, no placebo, parallel-group; second randomization: sotalol vs. propafenoneParoxysmal or persistent AF in the previous 6 months; minimum documented AF duration > 10 minSinus rhythm or AF followed by DCC with 21 days after randomizationTime to first ECG documented recurrence of AF > 10 minDay 21 was considered the start of follow-up (day 0); patients in AF on day 21 were classified as having had a recurrence on day 1; minimum follow-up 12 months; mean 16 monthsECG at 3, 6, 12, 18, 24 months; TTM during symptoms
SAFIRE-D, 2000325Dofetilide (3 doses)Double-blind, placebo-controlled, dose-ranging, parallel-groupPersistent AF for 2–26 weeksAF at enrollment in the in-hospital phase; sinus rhythm in order to enter the maintenance phase; DCC if no conversion after 5 doses of the drugIn-hospital: conversion to sinus rhythm; outpatient: time to first recurrence of AF or flutter for at least 24 hIn-hospital conversion phase (3 days/5 doses); outpatient maintenance phase; maximum follow-up 12 monthsIn-hospital: telemetry for 3 days; outpatient visits at 2 weeks, 1, 2, 3 months, 3-month intervals thereafter
Bellandi, 2001300Propafenone vs. sotalolDouble-blind, placebo-controlled, parallel-groupParoxysmal or persistent AF <48 h at enrollment; ≥4 episodes in the previous yearSinus rhythmTime to first symptomatic recurrence of any atrial arrhythmia; secondary endpoint: ventricular rates during first symptomatic recurrenceStabilization period (0–7 days) in patients on prohibited drugs (1–6 months for amiodarone); qualifying period (up to 28 days) to document AF > 10 min; loading period (4 days after randomization); efficacy period (up to 91 days)TTM or ECG during symptoms
ERAFT, 2002594Propafenone sustained release (2 doses)Double-blind, placebo-controlled, dose-ranging, parallel-groupParoxysmal, duration not specifiedSinus rhythmTime to first symptomatic recurrence of any atrial arrhythmia; secondary endpoint: ventricular rates during first symptomatic recurrenceStabilization period (0–7 days) in patients on prohibited drugs (1–6 months for amiodarone); qualifying period (up to 28 days) to document AF > 10 min; loading period (4 days after randomization); efficacy period (up to 91 days)TTM or ECG during symptoms
RAFT, 2003523Propafenone sustained release (3 doses)Double-blind, placebo-controlled, dose-ranging, parallel-groupParoxysmal, ECG documented in the previous 12 monthsSinus rhythmTime to first symptomatic recurrence of AF, flutter or SVT; secondary analysis in patients who were in the study on day 5; secondary endpoint: ventricular rates during first symptomatic recurrenceLoading period (1–4 days); maximum follow-up 39 weeksECG at 1, 3, 6, 12, 24, and 39 weeks; TTM at 2-week intervals and during symptoms
PAFAC, 20041182aSotalol, quinidine + verapamilDouble-blind, placebo-controlled, parallel-groupPersistent AF >7 daysSinus rhythm for at least 2 h after DCCTime to first recurrence of any AF or death; secondary endpoint: incidence of and time to persistent AFMaximum follow-up 12 months; mean follow-up 233 daysECG monthly; TTM daily and during symptoms
SOPAT, 20041033Sotalol, quinidine + verapamil (2 doses)Double-blind, placebo-controlled, parallel-groupParoxysmal; 1 episode in the previous monthSinus rhythmTime to first recurrence of symptomatic paroxysmal AF or drug discontinuationMinimum follow-up 12 months; extended follow-up beyond 1 year optional in patients without recurrence; mean follow-up 266 daysECG at 1 week and 1, 3, 6, 9, 12 months; 1-min TTM daily and during symptoms
SAFE-T, 2005665Amiodarone, sotalolDouble-blind, placebo-controlled, parallel-groupPersistent AF ≥72 hAF, DCC if no conversion on the drug within 28 daysTime to first recurrence after sinus rhythm was restoredMinimum follow-up 12 months, maximum follow-up 54 monthsECG monthly; TTM weekly; AF to be confirmed by 2 TTM recordings or TTM and ECG within 24 h
A-COMET I, 2006446AzimilideDouble-blind, placebo-controlled, stratified by the presence of structural heart diseasePersistent AF >48 h and <6 monthsAF, DCC on day 6 if no conversion on treatmentComposite endpoint: any AF, flutter, or SVT ≥ 24 h on day 4; or any AF, flutter, or SVT < 24 h leading to hospitalization or DCC; no sinus rhythm on day 4–6; withdrawal before day 4Loading period in-hospital (3 days), efficacy period (day 4–6 to maximum 26 weeks); patients who were cardioverted after first recurrence continued in the studyECG at 2, 4, 6, 8, 10, 12, and 26 weeks and during symptoms
A-COMET II, 2006658Azimilide, sotalolDouble-blind, placebo-controlled, parallel-groupPersistent AF >48 h and <6 monthsAF, DCC after 3 day loading period if no conversion (DCC considered successful if sinus rhythm was maintained for at least 1 h)Time to first recurrence of any AF, flutter, or SVT ≥ 24 h or AF, flutter, or SVT < 24 h leading to hospitalization or DCC; DCC failure on day 4; other endpoints: DCC facilitation, symptoms, conversion on treatmentLoading period with continuous monitoring (3 days), efficacy period (day 4 to maximum 6 months)ECG at 2, 4, 6, 8, 10, 12, and 26 weeks and during symptoms
FAFI, 2007143Flecainide (3 doses)Double-blind, placebo-controlled, dose-ranging, parallel-groupParoxysmal (at least 2 episodes), atrial flutter includedSinus rhythmThe absence of recurrent symptomatic AF or flutter; secondary endpoints: time to first recurrence of AF or flutter; number of symptomatic episodes, the absence of symptomatic and asymptomatic recurrence, duration of symptomatic AF or flutter, symptomsObservational period (4 weeks) to obtain at least 1 TTM showing AF >1 min; stabilization period (3 days); efficacy period (4 weeks)ECG at 4 weeks; 30 s TTM  ‘routinely’ (exact frequency not stated, probably daily) and during symptoms
EURIDIS and ADONIS, 20071237DronedaroneDouble-blind, placebo-controlled, 2:1 randomizationParoxysmal or persistent; atrial flutter includedSinus rhythm for at least 1 h; if needed, DCC before randomizationTime to first recurrence of AF > 10 min; secondary endpoint: time to first symptomatic recurrence of AF; mean ventricular rates during first recurrence12 monthsECG on days 7, 14, 21 and at 2, 4, 6, 9, 12 months; TTM (2 recordings 10 min apart) on days 2, 3, 5 and at 3, 5, 7, 10 months, and during symptoms
DIONYSOS, 2009504Dronedarone vs. amiodaroneDouble-blind, parallel-group, active comparatorPersistentAF > 72 h, DCC if required on day 10–28Time to first recurrence (including unsuccessful DCC) or drug discontinuation; secondary safety endpointsScreening period ( 30 days); DCC on day 10–28; maximum follow-up to 6 months after the last patient included ECG on days 1, 5, 10–28, 40 and at 3, 6,  9, 12 months, end of treatment, EOS
GISSI-AF, 20091442ValsartanDouble-blind, placebo-controlled, parallel-groupParoxysmal or persistent; at least 2 episodes in the previous 6 monthsSinus rhythm for at least 48 h before randomization, DCC between 14 days and 48 h before randomizationTime to first recurrence of AF and proportion of patients with >1 recurrence; secondary endpoints: number of AF episodes per patients12 monthsECG at 2, 4, 8, 24, 52 weeks and during symptoms; 30 s TTM weekly and during symptoms
POM-3, 2010663Prescription fatty acidsDouble-blind, placebo-controlled, parallel-groupParoxysmal (n= 542), persistent with at least 1 DCC (n= 121); at least 1 suspected or documented episode in the previous 3 months and at least 1 ECG documented episode in the previous 12 monthsSinus rhythmTime to first symptomatic recurrence of AF or flutter; secondary endpoint: time to first symptomatic recurrence of AF or flutter in patients with persistent AF and in the combined strataLoading period (1 week), maximum follow-up 6 monthsTTM bi-weekly and during symptoms
  • A-COMET, Azimilide–CardiOversion MaintEnance Trial; ADONIS, American-Australian-African trial with DronedarONe In atrial fibrillation or flutter for the maintenance of Sinus rhythm; AF, atrial fibrillation; AFL, atrial flutter; CTAF, Canadian Trial of Atrial Fibrillation; DCC, direct current cardioversion; ECG, electrocardiogram; EMERALD, European and Australian Multicenter Evaluative Research on Dofetilide; ERAFT, European Rythmol/rythmonorm Atrial Fibrillation Trial; EURIDIS, EURopean trial In atrial fibrillation or flutter patients receiving Dronedarone for the maIntenance of Sinus rhythm; FAFI, Flecainide Atrial Fibrillation Investigators; GISSI-AF, Gruppo Italiano per lo Studio della Sopravvivenza nell'Insufficienza cardiaca Atrial Fibrillation; PAFAC, Prevention of Atrial Fibrillation After Cardioversion; RAFT, Rythmol Atrial Fibrillation Trial; SAFE-T, Sotalol Amiodarone atrial Fibrillation Efficacy Trial; SAFIRE-D, Symptomatic Atrial Fibrillation Investigative Research on Dofetilide; SOCESP, Cardiology Society of São Paulo Investigators; SOPAT, Suppression Of Paroxysmal Atrial Tachyarrhythmias; SVT, supraventricular tachycardia; TTM, transtelephonic monitoring.

  • aEight hundred and forty-eight randomized.

While time to first recurrence is relatively easily measured, this outcome parameter may lead to excluding a substantial proportion of patients through the clustering phenomenon who may potentially benefit from continuous therapy in terms of the frequency and duration of subsequent recurrence, severity of symptoms, and the need for intervention such as repeat cardioversion. Conversely, several studies which utilized trans-telephonic ECG monitoring, have reported that, although antiarrhythmic drugs often fail to abolish all AF recurrences, they may render AF less symptomatic, less frequent, and less sustained,17,30,31 and as a result, may lead to improvement in patient's quality of life. This can be a clinically relevant effect adjuvant to, but not substituting the time to first recurrence.

It is possible that an antiarrhythmic drug may have a different effect in paroxysmal and persistent AF. For example, previous trials with dofetilide suggested that the drug was effective in preventing persistent AF, but had relatively little effect on the paroxysmal arrhythmia. Such a result might be due to differential antiarrhythmic effects on near-normal atria and on atria exposed to substantial remodelling—this phenomenon has been demonstrated experimentally.32 It is reasonable to perform small studies in paroxysmal and persistent AF separately before contemplating larger efficacy and safety trials. There is also an issue of appropriate follow-up duration (1–3 vs. 6–12 months), the requirement for a blanking period (e.g. for dose titration and stabilization) as is virtually routine in studies of left atrial ablation. However, any delay between randomization and the count of efficacy endpoints (safety endpoints can never be ‘blanked out’) introduces opportunities for post randomization events to influence the quality of the randomization.

Because antiarrhythmic drugs render some AF episodes asymptomatic, ECG monitoring is essential. A major issue related to the use of the ECG as a measure of outcome revolves around the means and duration of monitoring (e.g. Holter monitoring of various duration, external loop recorders up to 28 days, trans-telephonic monitoring, and implantable rhythm control and rhythm monitoring devices).13 It has been shown that intensified rhythm monitoring with implantable ECG recorders (e.g. Medtronic Reveal XT, St Jude Medical Confirm) with improved monitoring capabilities and extended memory, may yield a high diagnostic accuracy and also ensure adequate assessment of AF burden.33 Notably, such devices may potentially provide reliable information on the proarrhythmic safety of antiarrhythmic drugs.

Atrial fibrillation burden

Time to first recurrence may not provide the ideal information to assess antiarrhythmic efficacy in paroxysmal AF because of the non-random distribution of recurrences. Recurrences of paroxysmal AF do not occur in a completely random fashion described by a Poisson distribution, but tend to cluster at a period of time, better fitting a Weibull distribution.13 Counting the number and duration of AF events is an attractive measure of overall AF burden, but it is not devoid of significant limitations. Antiarrhythmic drugs, in addition to rendering some AF recurrences asymptomatic, may disrupt the AF pattern by ‘breaking’ typically prolonged episodes into several episodes of shorter duration. Counting the number of events in this case may lead to a false conclusion of a lack of efficacy or even a ‘proarrhythmic’ effect of the drug, whereas from the clinical standpoint short, asymptomatic AF episodes may be less clinically relevant, provided that rate control and risk factors have been appropriately addressed.

The use of implantable rhythm monitoring devices to log all events and calculate AF burden on different doses of an investigational compound, active comparator, or placebo is an attractive trial design (Table 3). An example of such a trial is a recent dose-ranging PASCAL (Paroxysmal Atrial Fibrillation Study with Continuous Atrial Fibrillation Logging) study of an amiodarone-derivative budiodarone which used the change in AF burden as a primary endpoint (presented at the Heart Rhythm Society 2009 Scientific Sessions). Dual-chamber pacemaker diagnostics were used to log all arrhythmic events during 12 weeks of treatment with three doses of budiodarone or placebo, and 4-week periods of baseline and wash-out. The study demonstrated a clear dose-dependent reduction in AF burden associated with treatment. The methodology also allowed assessment of several secondary endpoints such as number and duration of AF episodes, and duration of sinus rhythm.

View this table:
Table 3.

Antiarrhythmic drug trials using atrial fibrillation burden as a primary endpoint

StudyNumber of patientsDrugDesignType of AFDefinition of AF burdenAF monitoringSecondary endpointsDuration of the study
J-RHYTHM II, 2011318Candesartan vs amlodipineOpen label, no placeboParoxysmal, within 6 monthsNumber of days with documented AF per month (difference between the observation period and the final month)30-s TTM daily and during symptomsProgression to persistent AF > 7 days or requiring cardioversion; MACE a; LA size; QoLObservation period (4 weeks), maximum follow-up 12 months
ANTIPAF (NCT 00098137), reported 2010425OlmesartanDouble-blind, placebo- controlled, stratified by the use of beta-blockersParoxysmal, at least 1 episode in the previous 6 monthsPercentage of days with documented episodes of paroxysmal AF (number of days with AF divided by number of days with > 1 readable TTM-ECG)60-s TTM daily and during symptomsTime to 1st recurrence; time to persistent AF; time to prescription of the recovery drug b; time to 1st symptomatic recurrence; percentage of days with AF after 90 days of therapy; QoL; cardiovascular hospitalization; unscheduled outpatient visits for cardiovascular reasons; cerebrovascular events12 months
ARYx, 20096Budiodarone (ATI-2042)Open label, dose- escalating, no placeboParoxysmal, with AF burden 1–50% (mean, 20.3 + 14.6%; range, 4.6–45.3%)Percentage of time in AF (time in AF divided by the total time in each study period)EGM data downloaded from a dual-chamber pacemaker on days 8 and 14 of each 2-week periodNumber and duration of AF episodes; safety12 weeks in total: 4 dose-escalating periods, no treatment during baseline and washout (each period 2 weeks)
PASCAL (NCT 00389792), reported 200972BudiodaroneDouble-blind, placebo- controlled, parallel-groupParoxysmal, with AF burden > 3% in the previous 30 days (mean 26%; range, 1– 86%)Percentage of time in AF (change from baseline over 12 weeks of treatment compared with placebo)EGM data from a dual-chamber pacemakerNumber and duration of AF episodes; duartion of sinus rhythm; symptoms20 weeks in total: baseline and follow- up (washout) periods (4 weeks each), treatment period (12 weeks)
HESTIA (NCT 01135017), ongoing430DronedaroneDouble-blind, placebo- controlled, parallel-groupParoxysmal AF or flutter within 6 months; AF burden ≥ 1% on pacemaker interrogation at the screening visit with ≥ 1 AF episode within 28 daysPercentage of time in AFEGM data from a dual-chamber pacemakerNumber (per day) and duration (minutes) of AF episodes; ventricular rates; QoL; correlation between AF burden and symptoms; number of cardioversions or overdrive pacing; progression to persistent AF; safety12 months
NCT 01356914, planned20BMS-914392Double-blind, placebo- controlled, 4- way crossoverParoxysmal AF, with AF burden 1–50% on pacemaker interrogation at screeningNot specifiedEGM data from a dual-chamber pacemakerNumber and duration of AF episodes; ventricular rates; QoL; safety12 weeks (each crossover period 3 weeks)
  • AF, atrial fibrillation; ANTIPAF, angiotensin II ANTagonist In Paroxysmal Atrial Fibrillation; ECG, electrocardiogram; EGM, electrogram; HESTIA, Effects of Dronedarone on Atrial Fibrillation Burden in Subjects with Permanent Pacemakers; J-RHYTHM II, Japanese Rhythm Management Trial for Atrial Fibrillation; LA, left atrium; MACE, major adverse cardiovascular events; PASCAL, Paroxysmal Atrial fibrillation Study with Continuous Atrial fibrillation Logging; QoL, quality of life; TTM, transtelephonic monitoring.

  • a Cardiac death, myocardial infarction, cerebral infarction, heart failure or major bleeding requirung hospitalization

  • b Amiodarone

However, AF burden as a primary outcome measure has several technical and patient-related limitations (Table 4). The relationship between AF burden and traditional ECG endpoints in paroxysmal AF and major cardiovascular outcomes has not been established. Another technical disadvantage of AF burden as an outcome parameter in paroxysmal AF is the potential exclusion of patients whose AF burden at baseline is not sufficiently high and on the other hand, inclusion of those who have been picked up during the cluster of AF events. Similarly, it is difficult to know how to handle patients who develop persistent AF during the study since its duration (and hence ‘burden’ depends on the physician's choice of when to terminate the arrhythmia. If AF burden is used as an outcome measure in paroxysmal AF, a relatively extended monitoring period would be required.

View this table:
Table 4

Limitations of atrial fibrillation burden as an endpoint in antiarrhythmic drug trials

Atrial fibrillation episodes must occur often, or the study must be long
Only paroxysmal atrial fibrillation is suitable
Progression to persistent atrial fibrillation is difficult to handle
An implantable rhythm control/monitoring device (cardioverter–defibrillator, pacemaker or loop recorder) with adequate atrial sensing and must be available
Over-sensing (far-field) or under-sensing (blanking) may occur
Automatic detection of atrial tachyarrhythmias is essential
Sufficient memory must be available to prevent over-writing
Atrial or ventricular pacing may disturb the occurrence of the arrhythmia

Patients with paroxysmal AF and implanted dual-chamber pacemakers capable of detecting atrial arrhythmias may be an attractive population for such trials, but results from these very specific patients may not be generalizable to a broader population with paroxysmal AF. Such studies should be followed by larger and longer studies in wider patient populations where the use of less intensive monitoring such as trans-telephonic monitoring is more appropriate and practical (Table 3). Another way of overcoming the problem of variability of AF burden, without resort to implantation of expensive devices, is continuous monitoring using an external recorder for up to 28 days followed by 1–2 months without monitoring after which 28-day monitoring can be repeated. This methodology allows AF burden to be measured and is currently used in several antiarrhythmic drug studies.

Endpoints in studies of drugs for pharmacological cardioversion

The major limitations of pharmacological cardioversion in clinical routine are the relatively unpredictable outcome of treatment (conversion of AF), a generally lower efficacy compared with electrical cardioversion, limitations of drug choice because of underlying heart disease, and the risk of proarrhythmia.

The claimed indication for this group of drugs (usually intravenous formulations) is restoration of sinus rhythm, and the primary endpoint of clinical trials is termination of AF or atrial flutter (successful conversion to sinus rhythm) documented by continuous ECG monitoring. There are no strict criteria for the duration of sinus rhythm after conversion of AF following drug administration, which should be regarded as clinically relevant. While it is reasonable to define successful cardioversion simply as termination of AF (conceptually for only one beat), the clinical consequences of immediate recurrence of AF, after seconds or minutes of sinus rhythm, are not different from failed cardioversion. Ideally, the proportion of patients converted to sinus rhythm, subsequent time to recurrence and recurrence rates should be reported in an antiarrhythmic drug trial. The choice of the primary outcome depends on the therapeutic intention: if a drug is specifically developed for cardioversion of AF, time to conversion to sinus rhythm is a relevant outcome even if the duration of sinus rhythm is short and the clinical utility of the drug can be included within a more comprehensive strategy of restoration and maintenance of sinus rhythm. A drug that is ultimately intended to maintain sinus rhythm over long periods of time may, in contrast, be more useful when it also suppresses immediate (within 5 min), intermediate (6 min to 28 days), and possibly late (>28 days) recurrences of AF.12

Endpoints in studies of drugs for ventricular rate control

There is limited information available on the optimal target heart rate during AF. Few systematic studies explored the effect of rate slowing drugs on chronotropic competence in AF or define upper limits of the appropriate ventricular rate response during exercise. The definitions of optimal rate control are rather arbitrary. In retrospective analysis of the AFFIRM and RACE (RAte Control vs. Electrical cardioversion) studies, ventricular rates <100 bpm were associated with a better outcome compared with ventricular rates above this threshold.34 The RACE (RAte Control Efficacy in permanent atrial fibrillation) II study, conducted in a prospective, randomized, open-label fashion, showed no difference in the proportion of patients with a primary outcome event between strict rate control defined as a ventricular rate <80 bpm at rest and <110 bpm during moderate exercise and less rigorous rate control (ventricular rates at rest < 110 bpm).35

In the context of trials with new rate-controlling substances, reductions in the mean ventricular rate during rest and at a level of activity that can be achieved by the patients (e.g. assessed during an exercise test performed at 25% of maximum exercise)35 appears to be a reasonable outcome parameter. Alternatively, the mean heart rate during 24 h Holter monitoring can be used, with inclusion of the minimum and maximum ventricular rates and other Holter-derived parameters as secondary endpoints. Continuous detection of ventricular rates by implanted devices is likely to unravel a significantly higher incidence of episodes with uncontrolled ventricular rates than standard methods.36,37

Indirect evidence points at a modest improvement in exercise tolerance and left ventricular performance with appropriate rate control.38,39 Therefore, it is reasonable to measure the clinical benefit of agents with an effect on ventricular rates by assessing the symptomatic status, left ventricular function, exercise tolerance, quality of life, and morbidity and possibly mortality as outcomes.

Atrial fibrillation-related secondary endpoints

Quality of life

Further clinically relevant endpoints include quality-of-life measures with existing and potentially new AF-specific questionnaires, the EHRA symptom score, left ventricular function, left atrial function, cognitive function as well as health economics. Although symptom relief and improvement in quality of life are major therapeutic goals, they are currently recommended as secondary outcome parameters in clinical trials in AF because there are no reliable tools to quantify these parameters objectively and the effect of treatment cannot be assessed sufficiently reliably to support drug registration.13 There is considerable discrepancy between reported symptoms and the documented rhythm, and therapies can further alter this relationship.17,30,31,40,41 Therapies rendering AF asymptomatic may result in improvement in quality of life.42 However, even patients with seemingly asymptomatic AF report lower quality of life on some scales of SF-36 compared with counterparts without AF.43

Impaired quality of life may be secondary to the underlying heart disease which may affect the individual perception of health. Until recently, there have been no specific AF-oriented quality-of-life questionnaires to extract the effect of AF on quality of life. Beyond their direct effects on AF-related symptoms, interventions, including drug therapy, can also affect quality of life indirectly due to the potential side effects and the potential requirement for additional tests. Assessment of quality of life may become a vital tool to differentiate between treatment options, particularly, when patient preferences are concerned.

Recently, several disease-specific quality-of-life questionnaires have been proposed.4447 All these require validation and, when used in antiarrhythmic drug trials, should be used in conjunction with established short instruments such as EQ5D or SF-36 and the EHRA score. The EHRA symptom score system5 was developed specifically to simplify and standardize the assessment and quantification of AF-related symptoms in clinical practice and clinical trials. The EHRA classification relates specifically to the time when the patient feels to be in AF and provides a specific quantification of the symptoms that are attributable to the functional consequences of AF. The system will require validation, but appears a simple and easily available tool for assessing AF-related symptoms in trials. As a secondary endpoint in drug trials, measurement of quality of life is equally useful in patients with all forms of AF, but is probably a more effective tool in patients with paroxysmal AF who typically exhibit higher levels of quality-of-life impairment and therefore, will show more benefit from therapy.48 Assessment of quality of life is essential when the treatment is claimed to suppress symptomatic AF.

Left ventricular function and exercise tolerance

Atrial fibrillation can impair left ventricular function via multiple mechanisms, and left ventricular dysfunction can also predispose to AF. There is evidence from observational and some rate vs. rhythm control and observational studies pointing at a modest reduction in the left ventricular and atrial sizes and improvement in left ventricular systolic function with the rhythm control strategy.39,49,50 Successful ablation of AF has also been reported to improve impaired systolic ventricular function.51 Conversely, optimization of ventricular rate control in patients with suspected tachycardia-induced cardiomyopathy may lead to a reversal of ventricular impairment.38 Therefore, it is reasonable to use different measures of left ventricular performance, such as serial echocardiograms to measure left ventricular volumes or dimensions, and left ventricular ejection fraction systolic function as outcome measures in studies of agents with an effect on the ventricular rates.

This can be further supplemented by an exercise stress test as a general assessment of cardiac performance. Some studies reported a sustained 15% improvement in exercise tolerance with restoration and maintenance of sinus rhythm.52 However, the large inter-individual variability in patient response has not been fully explained, although several factors, such as age, exercise tolerance at baseline, and obesity, are seen as potential determinants of a poor response. Patients with recurrent AF demonstrated a similar improvement in exercise tolerance when tested during sinus rhythm as did patients who remained in sinus rhythm. In summary, this parameter is only moderately reliable, particularly in patients with non-permanent AF.

Need for active comparator studies

From the regulatory standpoint, the clinical trials of drugs for restoration and maintenance of the sinus rhythm or rate control should ideally include placebo and an active comparator. The choice of an active comparator in cardioversion studies is determined by the characteristics of the target population and the availability of alternative therapy. In rhythm maintenance studies, the choice often depends on underlying disease state.

In general, the European Medicines Agency requires studies with an active comparator.12 The use of placebo is necessary to determine effect size which subsequently should be considered in the clinical context by conducting an additional comparison with a therapy which is already available. Three arm studies including placebo can be considered to ensure assay sensitivity. An active comparator will also be required when the use of placebo is unethical.

Although active controlled studies are feasible and the use of an active comparator offers an advantage of a long-term comparison of the efficacy and safety, puts the investigational drug in the clinical context, and also allows assessment of clinical outcomes, there are several limitations to this study design. A suitable comparator, particularly with a known effect on mortality and morbidity, may not always be available, and the results may be affected by the limitations of approved indications and the populations studied with the reference compound. There are methodological issues of assay sensitivity and non-inferiority and the clinical implications in terms of outcome are unclear when surrogate endpoints are used.


The clinical development of new antiarrhythmic drugs in AF is of high clinical and socio-economic relevance. The complexity of AF itself and of its cardiovascular consequences, as well as the diverse and often elderly and comorbid patient populations affected by AF, require careful selection of patients and clear definition of measurable and relevant clinical outcomes within the clinical development program of a new antiarrhythmic drug. ECG-based outcomes are indispensable for an antiarrhythmic drug, but require supplementation by measurements of mortality—at least with the intention to rule out harmful effects—and AF-related morbidity such as stroke, acute heart failure, and myocardial infarction. Safety is a major concern in the clinical use and in the development and approval process of new antiarrhythmic drugs for AF. The effect of a new drug in the different outcome domains (death, stroke and other cardiovascular complications, quality of life, rhythm) and the study populations will affect the labelling and clinical used new antiarrhythmic drugs in AF.

Conflict of interest: I.S. is an advisor/speaker/investigator for Sanofi, Bristol-Myers Squibb, Takeda, Daiichi, Boehringer Ingelheim, Servier, Astellas, Mitsubishi Pharma, and Merck. P.K. has received consulting fees and honoraria from 3M Medica, AstraZeneca, Bayer Healthcare, Boehringer Ingelheim, MEDA Pharma, Medtronic, Merck, MSD, Otsuka Pharma, Pfizer/BMS, sanofi-aventis, Servier, Siemens, and TAKEDA; research grants from 3M Medica/MEDA Pharma, Cardiovascular Therapeutics, Medtronic, OMRON, sanofi-aventis, St. Jude Medical, the German Federal Ministry for Education and Research (BMBF), Fondation Leducq, the German Research Foundation (DFG), and the European Union (EU); and travel grants from the European Society of Cardiology (ESC), the European Heart Rhythm Association (EHRA), a registered branch of the ESC, and from the German Atrial Fibrillation Competence NETwork (AFNET). N.D. is an advisor/speaker/investigator for Astra Zeneca, Bristol-Myers, Squibb, Merck Sharp & Dohme, Novartis, Pfizer, sanofi aventis. A.J.C. is an advisor/speaker/investigator for Servier, Novartis, Sanofi, Astra Zeneca, Cardiome, Prism, Astellas, Xention, ARYx, Prism, Bristol-Myers Squibb, Daiichi, Merck, Boehringer Ingleheim, Medtronic, St Jude Medical, Biotronic, and Boston Scientific.


A.J.C. is supported by the British Heart Foundation.


  • Report on the meeting between representatives of the European Society of Cardiology, the pharmaceutical, medical device and imaging industry, and European Medicines Agency (EMA) regulators held in London on 15–16 September 2009.


View Abstract