OUP user menu

Outcome parameters for trials in atrial fibrillation: executive summary

Paulus Kirchhof, Angelo Auricchio, Jeroen Bax, Harry Crijns, John Camm, Hans-Christoph Diener, Andreas Goette, Gerd Hindricks, Stefan Hohnloser, Lukas Kappenberger, Karl-Heinz Kuck, Gregory Y.H. Lip, Bertil Olsson, Thomas Meinertz, Silvia Priori, Ursula Ravens, Gerhard Steinbeck, Elisabeth Svernhage, Jan Tijssen, Alphons Vincent, Günter Breithardt
DOI: http://dx.doi.org/10.1093/eurheartj/ehm358 2803-2817 First published online: 25 September 2007


Atrial fibrillation (AF), the most common atrial arrhythmia, has a complex aetiology and causes relevant morbidity and mortality due to different mechanisms, including but not limited to stroke, heart failure, and tachy- or bradyarrhythmia. Current therapeutic options (rate control, rhythm control, antithrombotic therapy, ‘upstream therapy’) only prevent a part of this burden of disease. Several new treatment modalities are therefore under evaluation in controlled trials. Given the multifold clinical consequences of AF, trials in AF patients should assess the effect of therapy in each of the main outcome domains. This paper describes an expert consensus of required outcome parameters in seven relevant outcome domains, namely death, stroke, symptoms and quality of life, rhythm, left ventricular function, cost, and emerging outcome parameters. In addition to these ‘requirements’ for outcome assessment in AF trials, further, more detailed outcome parameters are described. In addition to a careful selection of a relevant primary outcome parameter, coverage of outcomes in all major domains of AF-related morbidity and mortality is desirable for any clinical trial in AF.

  • Atrial fibrillation
  • Controlled trial
  • Outcome parameter
  • Therapy, treatment
  • Randomized trial
  • End-point
  • Stroke death
  • Quality of life
  • Left ventricular function
  • Catheter ablation
  • Antiarrhythmic drugs
  • Cardioversion
  • Rate control
  • Rhythm control
  • Anticoagulation


Atrial fibrillation (AF) affects a relevant, increasing part of the population of the European Union:1 25% of the currently 40-year-olds will develop AF.2,3 AF causes relevant mortality and morbidity.47 The clinical syndrome ‘AF’ includes a broad spectrum of pathophysiological processes, ranging from ‘electrical accidents’8,9 to long-term systemic processes.10 Likewise, the clinical consequences of AF are difficult to predict in an individual patient. Death, stroke, or severe limitations of exercise capacity in some patients contrast with frequent asymptomatic AF episodes in others.1113 Treatment includes antithrombotic medicines,14 control of ventricular rate by drugs or pacemakers,15 rhythm-control interventions,1618 and the so-called ‘upstream therapy’.19

The limited effectiveness and at times unfavourable side-effect profile of available therapeutic options has resulted in a massive surge of randomized trials in AF in the past two decades (Figure 1). Due to the diverse therapeutic options and desired outcomes, trials that assess different treatment options for AF often use completely different outcome parameters. This makes AF a difficult topic in the context of controlled trials and clinical day-to-day management.

Figure 1

Number of randomized trials in atrial fibrillation published in Medline from 1967 to 2006. Dots indicate the mean number of trials per year over a given decade and error bars indicate the standard deviation.

To tackle this problem, the German Atrial Fibrillation competence NETwork (AFNET, www.kompetenznetz-vorhofflimmern.de) and the European Heart Rhythm Association (EHRA www.escardio.org/ehra) convened 60 scientists and industry representatives at the European Heart House in Sophia Antipolis, France, for a consensus conference on 22–23 January 2007 to define minimal and reasonable outcome parameters for the assessment of AF in controlled clinical trials. This paper reports the summarized consensus reached during this conference. Seven relevant outcome domains are covered in a hierarchical structure: Death, stroke, symptoms and quality of life, changes in rhythm, left ventricular (LV) function and development of heart failure, health economics, and emerging outcome parameters.

Natural time course of AF, frequency, and timing of complications

AF is often a chronic, progressive arrhythmia. The first detected episode of AF is usually self-terminating or amenable to rhythm control interventions, mostly followed by intervals of sinus rhythm, interrupted by episodes of the arrhythmia (‘paroxysmal AF’). The distribution and duration of AF episodes slowly increases over time, but is rather clustered20,21 than random.22 This biological pattern renders AF burden a cumbersome outcome parameter in terms of statistical power (Figure 2).21,23 The recurrence of persistent AF, in contrast, is a single event and ‘time to recurrence’ is a reasonable outcome since this event often triggers a modification of therapy. After restoration of sinus rhythm, e.g. by cardioversion, persistent AF recurs in 25–50% of all patients in the first month after cardioversion.12,24 Thereafter, AF recurs in ∼10% of patients per year (Table 1).

Figure 2

Time course of atrial fibrillation (AF). Shown is a typical chaotic pattern of time in AF (black) and time in sinus rhythm (grey) over time (X-axis). AF progresses from undiagnosed to first diagnosed, paroxysmal, persistent, and finally to permanent AF. Flashes indicate cardioversions as examples for therapeutic interventions that influence the ‘natural’ time course of the arrhythmia.

View this table:
Table 1

Points to consider on the natural time course of atrial fibrillation

AF tends to progress to permanent AF:∼10% in the first year after symptomatic manifestation, 5% per annum thereafter.
Structural heart disease and age may promote this progression.
Paroxysmal AF recurrences follow chaotic patterns that are not random.
Recurrence of persistent AF can be classified as immediate, early, and late. Recurrent AF is most frequent in the first weeks after cardioversion.
Presence of AF approximately doubles mortality, with a likely even higher impact on cardiovascular mortality
  • AF, atrial fibrillation.

AF causes a variety of complications, most notably thromboembolic complications and reduced cardiac performance. The frequency of AF-related complications depends on the baseline patient characteristics.25,26 Furthermore, complications vary over time: The risk of severe bleeding appears highest in the first months after initiation of antithrombotic treatment. Death is highest in the first year after the initial manifestation of AF,5,27 and occurs steadily thereafter at 1.6–4.2% per annum in more recent controlled trials.16,22,28 Moreover, complication rates have decreased over the years in AF trials.14

General considerations

Patient characterization

The inherent risk for recurrent AF and AF-related complications is heavily influenced by patient characteristics. There are published lists of such baseline data by ACC/AHA29 and the society of Thoracic Surgeons.30 The panel suggests to refer to them, especially to the more general ACC/AHA recommendations.29 Table 2 lists a set of required, minimal baseline characteristics for publication of an AF trial.

View this table:
Table 2

Minimal clinical parameters that should be given for baseline characterization of patients in an atrial fibrillation trial

Age, Gender
Type of AF (first detected, paroxysmal, persistent, permanent)
Duration of AF since first detection
Prior antiarrhythmic drug treatment
 Number of antiarrhythmic drugs tested
 Number of cardioversions
 Number of catheter ablations or surgical interventions
CHADS2 score
Prior antithrombotic treatment
 Duration of anticoagulation (vitamin K antagonists, other anticoagulant)
 Anti-platelet treatment (aspirin, clopidogrel, etc)
Symptoms due to AF
 Arrhythmia-related symptoms (EHRA score)
 Prior stroke/transitoric ischaemic attack
Heart failure indices
 New York Heart Association class
 Left ventricular ejection fraction
Treatment at enrolment
 Antiarrhythmic drugs
 Rhythm control drugs
 Rate control drugs
 Antihypertensive therapy (special report of angiotensin receptor inhibition is suggested)
 Other cardiac medication
Concomitant cardiac disease
  • These data should be collected at study entry.

  • AF, atrial fibrillation.

Choice of outcome parameters in AF trials

Different AF trials will require different primary outcome parameters, but the complex consequences of AF will require assessment of a variety of outcome domains in every trial (Table 3). Ideally, the primary outcome reflects the efficacy of treatment. Safety concerns about the study treatments should be reflected in the primary safety outcome parameter. Outcome parameters reflecting net benefit should usually be presented as secondary outcome parameters. Generally, objectively measured parameters are preferred (but see section on Symptoms). If outcome measures do not cover all main outcome domains, at times pivotal information may be disguised.

View this table:
Table 3

Examples for primary outcome parameters in prior and ongoing clinical trials in AF

Trial acronym/NameNumber of patientsPrimary outcome parameter
Rate vs. Rhythm control trials
 PIAF13252Symptom improvement defined as elimination of palpitations, dyspnoea, and shortness of breath
 HOT-CAFÉ98205Composite (death and MACCE)
 STAF99200Composite (death, embolic events, and others)
Rhythm control trials
 SAFE-T24450Time to persistent AF
 PAFAC12866Time to persistent AF
 CTAF100403Time to persistent AF
 SOPAT511033Time to symptomatic AF
 ATHENA (NCT0017478)4300Death or cardiovascular Hospitalization
 Flec-SL (NCT00215774,55)755Time to persistent AF
 ANTIPAF (NCT00098137)422Time in AF
 Brignole101137Development of permanent AF
 Madrid102154Time to persistent AF
 Ueng103145Time to recurrent AF
 Natale10461Atrial flutter, rehospitalization, quality of life
 Wazni10570Time to recurrent AF, hospitalizations, and QoL
 APAF106198Time to recurrent AF
 Oral10780Recurrent AF (assessment not specified)
 Karch108100Freedom from AF in 7-day Holter
 Oral109146Freedom from AF at 1 year FU
 Gaita110105Freedom from AF at 2 years FU
 AF-CHF1111450Cardiovascular mortality
 RAAFT NCT00392054400Time to recurrent AF (>30 s)
 GAP-AF196Time to recurrent AF
 AMICA216Change in LV function
 CABANA3000Total mortality
Rate control trials
 AIRCRAFT7199Cardiac function, exercise capacity
 FARFIC7066Quality of life, exercise capacity
 OPSITE11256Quality of life, exercise capacity
 RACE II113500Composite
 Farshi1512Rate increase during exercise
Antithrombotic treatment trials
 AFASAK114335thromboembolic complication
 BAATAF115420Stroke, death (not defined)
 SPAF I1161330Stroke or peripheral embolism
 SPINAF117572cerebral infarction
 EAFT381007Composite (death from vascular disease, any stroke, myocardial infarction, or systemic embolism)
 CAFA118187Composite (non-lacunar stroke, non-central nervous systemic embolism, and fatal or intracranial haemorrhage)
 AFASAK 2119677Stroke or a systemic thromboembolic event
 SPAF II120715stroke or systemic embolism
 SPAF III1211044Stroke or systemic embolism
 SPORTIF III283410Stroke or systemic embolism
 SPORTIF V1223922Stroke or systemic embolism
 NASPEAF1231209Composite (vascular death and non-fatal stroke or systemic embolism
 TIARA (NCT00224757)300Composite (death, stroke, embolism, acute coronary syndrome, and major bleeding)
 ACTIVE W1246706Composite (MACE)
  • The panel has had difficulties to identify the primary outcome parameter in some of the published trials. The panel strongly recommends that the primary outcome parameter should be specifically stated in the publication of a trial.

  • AF, atrial fibrillation.

Assessment of specific outcome parameters


AF is associated with increased and premature mortality.4,5 Furthermore, many available treatment modalities, e.g. but not limited to antithrombotic medicines, antiarrhythmic drugs, catheter interventions, or operations, will at times cause death as a serious adverse event (SAE). All deaths therefore need to be measured and reported in any trial of AF on an ‘intention-to-treat’ basis from the time of randomization. There are several causes of death (Table 4). Deaths should be classified according to the mode of death using all available methods, including autopsy, doctors' reports, read-out of ICDs/monitoring devices, or Holter ECG recordings. Death unrelated to AF (e.g. death due to cancer) will dilute the effect of any treatment aimed at reducing AF-related mortality in a controlled trial. This effect will be more prominent in elderly study participants who are prone to die. Mortality should only be part of the primary outcome parameter when the therapy or intervention tested is aimed at reducing mortality and the trial has sufficient statistical power and sufficient follow-up time to detect an effect on mortality. This will only be possible in large trials that enrol patients at relatively high risk for death who are followed for a sufficient time. In short-term studies, studies in patients at low risk of death, and in studies in which the intervention will not affect mortality to a relevant extent, death is not a reasonable primary outcome parameter. Death should always be assessed as a secondary outcome parameter in such trials. Generally, death from unrelated causes should not be included in the primary outcome parameter unless the study is adequately designed to detect an effect of therapy on total death. However, all deaths should be reported as a safety outcome parameter.

View this table:
Table 4

Classification of deaths in atrial fibrillation trials

Non-cardiovascular, excluding sudden death
Cardiovascular death
 Sudden (including arrhythmic, myocardial infarction, among others)
 Vascular (e.g. embolic, subarachnoidal bleeds, stroke, other)
Treatment- or procedure-related (is also a serious adverse event)
  • All-cause death should be classified in the following groups.

  • AF, atrial fibrillation.

AF-related death is conceptually an attractive outcome parameter for AF trials, because it implies that the effect of the arrhythmia on mortality is directly measured. AF-related death should not substitute ‘total death’ as an outcome parameter, because AF-related death will be difficult to assess in a clinical trial, rendering AF-related death a potentially unreliable measurement, similar to and even more pronounced than cardiovascular death. Furthermore, there are no validated means to determine AF-related death. The panel acknowledges the potential relevance and the shortcomings of this outcome parameter and suggests a step-wise exclusion process to determine ‘AF-related death’: All deaths without a clearly determined non-cardiovascular cause should be classified as cardiovascular deaths. All cardiovascular deaths that do not have a clearly defined other cause (e.g. rupture of an aneurysm, pulmonary embolism, cardiac tamponade, myocardial infarction, among others) should be classified as AF-related death when AF was present during the 7 days prior to death. All deaths that are a consequence of AF-related treatment (SAE) should be reported in the primary safety outcome and counted as AF-related deaths. This process to determine AF-related death requires validation in prospective trials.


  • Mortality is a valid outcome parameter in AF trials when trials are adequately powered and designed to detect differences in mortality between treatment groups.

  • In the majority of trials, death is not a feasible primary outcome parameter, but may be part of a composite outcome parameter when the study treatment is aimed at reducing deaths.

  • Death is a required secondary outcome parameter. All deaths should be reported on an intention-to-treat basis, and information on vital status needs to be assessed at regular intervals (minimum: at enrolment and at the end of the trial).

  • All deaths must be reported in a safety outcome parameter.


AF causes a relevant portion of all strokes (15–25%)31, and AF-associated mortality is in part attributable to stroke and its consequences.32 Strokes in patients with AF are more severe than other forms of stroke.33 ‘Silent stroke’ is associated with AF and can be seen by cerebral imaging. Epidemiological data have associated silent cerebral ischaemic events with dementia.34 Stroke is often caused by cardioembolism in AF patients, most frequently from the left atrial (LA) appendage.35 Even in controlled trials, the residual stroke rate on optimal antithrombotic treatment (vitamin K antagonists, target INR range 2–3) is relatively high (1.3% per year in individuals without prior stroke, 3% per year in individuals with prior stroke)3639. Therefore, stroke is one of the most important outcome parameters in AF trials. Stroke should be evaluated using the best possible methods (including MRI/CT brain imaging, assessing intensity of anticoagulation, severity of stroke, and neurological end result). All stroke events should be adjudicated by a committee that is usually blind to treatment/study arm.

Intracerebral bleed is the natural counterpart of ischaemic stroke in anticoagulated patients. All bleeding events need to be reported. Bleeds become more prevalent during supratherapeutic anticoagulation (INR>3.540). Risk factors for bleeds include age, typical cardiovascular risk factors, and presence of cerebral small vessel disease. A bleeding event is major when it is fatal; haemoglobin concentration falls by more than 2 g/dL; requires transfusion of whole blood cells or operation; affects areas of concern, e.g. retroperitoneal, intracranial, intraspinal, or intraocular; or results in treatment cessation. Other bleeding events are minor. Intracranial bleeds may be included in a composite stroke outcome because the combination of strokes and intracerebral bleeds can reflect the clinical benefit of antithrombotic treatment. Subdural or epidural haemorrhages are not strokes, but should be reported as SAEs, together with a statement whether they appear attributable to treatment. Data on transient ischaemic attacks (TIAs) with acute lesion matching the symptoms on imaging should be collected and reported, as there is discussion on the classification of such outcome events, and a new definition that might classify such events as ‘stroke’ is under consideration at the World Health Organization. TIAs should always be adjudicated for the presence of stroke, and the clinical adjudication may determine the ultimate classification as stroke or TIA. Cause of stroke should be classified according to TOAST criteria, stroke end results by the Rankin score (Table 5, 41,42). Usually, cerebral vascular events and other major cardiovascular events (e.g. myocardial infarction, pulmonary embolism) should be assessed as separate outcome parameters.

View this table:
Table 5

TOAST criteria for classification of strokes (modified from Adams et al.41), and Rankin score for stroke severity (modified from Rankin42)

TOAST criteria: etiology of ischaemic strokes can be classified into five categories by clinical and imaging criteria
 Large-artery atherosclerosis
 Small-vessel occlusion
 Stroke of other determined etiology (e.g. large vessel dissection)
 Stroke of undetermined etiology.
Rankin score of stroke severity as described in Rankin42
 Grade I. No significant disability: able to carry out all usual duties.
 Grade II. Slight disability: unable to carry out some of previous activities but able to look after own affairs without assistance.
 Grade III. Moderate disability: requiring some help but able to walk without assistance.
 Grade IV. Moderate severe disability: unable to walk without assistance and unable to attend to own bodily needs without assistance.
 Grade V. Severe disability: bedridden, incontinent, and requiring constant nursing care and attention.
  • In larger studies, each TOAST category comprises ∼20% of all strokes. Cardioembolic strokes are often due to atrial fibrillation.

Cerebral imaging should be used to identify baseline cerebral defects in patients who suffered a cerebrovascular accident prior to trial inclusion. The preferred method of imaging is magnetic resonance imaging (MRI). Computed tomography may substitute MRI in specific situations. To measure cognitive function in trials, we recommend the Mini-Mental State Examination at baseline and during follow-up. Additional psychometric tests may be reasonable when cognitive function is part of the outcome. A minimal requirement is assessment of mini-mental state at enrolment and at the end of trial.


  • All strokes (ischaemic and haemorrhagic) and systemic embolic events should be recorded and reported separately.

  • All clinical events fulfilling the criteria of stroke should be verified by brain imaging, ideally by MRI.

  • TIAs are not counted as a stroke and should not be used as part of the stroke outcome.

  • Major bleeding should be reported separately, usually as a safety outcome parameter.

  • To assess cognitive function in trials, the Mini-Mental State examination should be recorded at baseline and at least also at the end of follow-up. If cognitive function is an outcome parameter, additional psychometric tests are recommended.

Symptoms and AF-related quality of life

AF is associated with poor quality of life.43 Rate- or rhythm control interventions can improve quality of life in AF patients.13,44,45 Symptoms and perceived suffering from the arrhythmia are the most common reason for AF patients to seek medical attention, and the main indication for rate- or rhythm-control therapy at present.35,46 Quality of life, and suffering or ‘illness intrusiveness’47,48 are difficult to measure objectively. The elusive relation between symptoms and arrhythmia recurrences, and specifically the high incidence of asymptomatic AF recurrences in patients with symptomatic AF, suggest that symptoms may at times not be related to AF, but rather an expression of other disease-causing processes. This renders symptoms and disease-related quality of life a potentially unreliable outcome parameter in AF trials. Therefore, symptoms and quality of life are only recommended as secondary outcome parameters.

Several instruments have been used to measure AF-related quality of life, usually as self-administered questionnaires [e.g. SF 36, symptoms check list, atrial fibrillation symptoms scale (AFSS), and the living with heart failure questionnaire (LWHF)]. These instruments are validated for global illness intrusiveness, but are—with the exception of the AFSS—not specific for AF-related symptoms. Furthermore, they are not available in many languages. Such standard instruments are recommended in AF trials, but the authors acknowledge their shortcomings in assessing AF-related symptoms. The panel recommends to design, validate, and use further, AF-specific instruments to assess AF-related quality of life, especially when improvement of symptoms and quality of life are the intended primary outcome of a trial. In selected studies in low-risk patients, robust, validated measures of quality of life may in the future become a primary outcome parameter, especially in small, hypothesis-generating trials.

Proposal of a symptom classification for AF

Having noticed the apparent discrepancy between the clinical relevance of AF-related symptoms for treatment decisions in AF and the lack of a practicable instrument to assess AF-related symptoms, the panel agreed to suggest a score to assess symptoms related to AF (see discussion in49). The panel suggests the so-called EHRA classification to describe AF-related symptoms (Table 6). The EHRA classification relates specifically to the time when the patient feels to be in the arrhythmia. The panel is aware of the fact that this classification requires prospective validation.

View this table:
Table 6

EHRA atrial fibrillation symptoms classification

Symptom severityDefinition
EHRA I‘no symptoms’
EHRA II‘mild symptoms’Normal daily activity not affected
EHRA III‘severe symptoms’Normal daily activity affected
EHRA IV‘disabling symptoms’Normal daily activity discontinued
  • The following items during presumed arrhythmia episodes are checked to determine the score: palpitations, fatigue, dizziness, dyspnoea, chest pain, anxiety. In addition to this score, the frequency could be classified into three groups, namely occasionally (less than once per month), intermediate (once per month—almost daily), and frequent (at least daily).


  • Symptoms are the main reason for AF patients to seek medical attention. At present, symptoms and quality of life are recommended as secondary outcome parameters because there are no reliable instruments to quantify AF-related symptoms.

  • Symptoms and quality of life should be assessed at entry and during follow-up in all AF trials.

  • In trials enrolling symptomatic patients, symptoms should be related to the underlying rhythm.

  • When the tested intervention is expected to primarily affect symptoms and quality of life, measures of quality of life and symptoms are potentially the primary outcome parameter. In such studies, the design, validation, and use of ‘specific’ instruments for AF-related symptoms in addition to standard instruments is recommended.

Once validated, the suggested EHRA AF symptoms classification may be helpful to compare AF-related symptoms across trials and in clinical practice.

Assessment of rhythm and other ECG-based outcome parameters

ECG-based outcome measures have been used in almost all trials that assessed interventions for rhythm or rate control (Table 7). In the past, we have learned that AF often recurs without clinical signs or symptoms: ECG recordings triggered by symptoms will miss more than half of all AF episodes, even in symptomatic patients.12,38,50,51 To detect both symptomatic and asymptomatic AF recurrences, systematic (scheduled) ECG recordings are therefore needed. Continuous ECG monitoring, the gold standard for detection of recurrent AF, is not available at present, and will be available only using advanced technology in the foreseeable future.

View this table:
Table 7

ECG-based outcome parameters for atrial fibrillation trials and available methods to assess them

ECG-based outcome parameters
 Freedom from AF (suitable for time-based assessment)
 Change in AF pattern (e.g. altered AF burden, altered AF type, among many others)
 Proarrhythmia (e.g. sudden death, ventricular tachycardia, torsades de pointes, atrial flutter, bradycardia, AV nodal block)
 Ventricular rate during AF at rest and during exercise
Available ECG methods
 Non-continuous standard ECG recording
  Symptom-activated ECG (e.g. during triggered visits, patient-activated devices)
  Algorithm-activated (device monitors rhythm)
   Resting ECG
   Transtelephonic monitoring
   (24–168 h) Holter recording
   Loop recorders
 Continuous ECG monitoring
  Pacemakers/implanted defibrillators
  ECG garment equipped with radio data transmission (e.g. GSM-based)
  • AF, atrial fibrillation.

In all patients, AF should be documented by ECG at enrolment, persistent, or permanent AF by Holter ECG. Seven day Holter ECG recording and daily plus symptom-activated transtelephonic ECG monitoring are equally powerful to detect recurrent paroxysmal AF5154 and detect ∼70% of AF recurrences. One has to accept that the negative predictive value for ‘freedom from AF’ is 25–40% in paroxysmal AF patients with the aforementioned monitoring intensity, indicating that only one in three patients without any detected AF in all monthly Holter ECGs or daily transtelephonic monitoring during a 1-year follow-up period will really be free of AF.23,53

On the basis of this knowledge, we suggest the following: All ECG recordings should be analysed blind-to-treatment in a core laboratory. Every perceived (symptomatic) episode of AF should trigger an ECG. Single-lead ECGs are sufficient for this monitoring. Additional scheduled regular ECG recordings are mandatory, either by scheduled 24 h/month Holter ECG, or by daily 30–60 s short-term (e.g. transtelephonic) ECG recordings.21,23,5153 Daily transtelephonic monitoring may be more feasible because it allows recordings of additional ECGs during times of perceived symptoms. Holter ECG recordings, in contrast, have the advantage that the duration of AF episodes can be assessed.

When the outcome parameter is freedom from persistent or permanent AF, daily or twice-weekly short-term (e.g. transtelephonic) ECG recording followed by rapidly scheduled confirmatory Holter recording in case of a documented AF recurrence are sufficient.12,24,55

Any arrhythmia that has the ECG characteristics of AF and lasts longer than 30 s should be reported as recurrent AF. Persistent or permanent AF is assumed to be present when the episode does not terminate spontaneously or is terminated by an intervention. In rhythm control trials, recurrent arrhythmias on drug or after catheter ablation will not always be AF, but at times constitute atrial tachycardias or atrial flutter. These should be included in the arrhythmia recurrence outcome parameter. Often, only a 12-lead ECG will distinguish these arrhythmias from AF.

Recurrent AF after cardioversion is time-dependent.5662 We suggest the following definitions: Cardioversion is successful when AF has been terminated. When AF recurs in the first 5 min after cardioversion, this event should be described as immediate recurrence of AF. AF recurrences within 6 min and 28 days after cardioversion should be called early recurrence of AF. Recurrent AF more than 4 weeks after cardioversion is ‘late’.6365


Safety issues may require additional ECG recordings to detect tachycardia and bradycardia signals. Bradycardia detection may require night-time ECG monitoring. Any arrhythmia that might constitute a proarrhythmic event (e.g. torsades de pointes, atypical or typical atrial flutter, or symptomatic bradycardia) must be reported as an adverse event.

Control of ventricular rate should be assessed by a resting ECG and a standardized submaximal exercise test (e.g. treadmill ECG, two-flights-of-stairs test, or a 6 min walk test). Maximal heart rate and potentially mean heart rate on Holter ECG may be used instead of an exercise test. Studies that compare rate- and rhythm-control strategies require detailed ECG monitoring like rhythm control studies (discussed earlier). In trials that do not target rate or rhythm, regular 12-lead ECGs should be performed, e.g. in 6 months intervals, to document the presence or absence of AF.


  • Every arrhythmia with the ECG characteristics of AF and a duration >30 s should be reported as an AF recurrence.

  • Every symptomatic event should trigger an ECG recording.

  • Regular scheduled additional ECG recordings are needed to detect asymptomatic episodes.

    • For detection of persistent or permanent AF, daily or twice-weekly short-term ECG recordings with rapidly ensuing confirmatory Holter in case of an arrhythmia recurrence are sufficient.

    • For detection of paroxysmal AF, regular Holter ECGs (24 h/month) or regular transtelephonic short-term ECGs (30–60 s once daily) are recommended. Even this intensity of ECG monitoring will not detect all patients with recurrent AF.

  • Ventricular rate should be assessed by resting ECG and a standardized exercise test. Alternatively, heart rate on Holter ECG may be used.

  • Safety measures may require additional ECG recordings, e.g. to detect proarrhythmia.

Left ventricular function and heart failure

AF may impair LV function. On the other hand, AF and its complications may also occur in the presence of LV dysfunction. Rate- or rhythm-control therapy can improve LV function.6673

LV size and function

LV function should be measured at baseline in every AF study patient.35 Echocardiography is widely available and provides real-time imaging. Long axis M-mode measurements assess LV size (LVEDD, LVESD) and estimate global LV function. M mode echocardiography will be sufficient in many trials in AF patients to give a global estimate of LV systolic function. When regionally or severely abnormal LV function is expected, apical 2-dimensional LV planimetry using a (modified) Simpson's approach is preferred.74 LV function should be measured at a normal heart rate (60–100 b.p.m.). In patients with irregular ventricular rhythm, averaging of LV measurements (five consecutive beats) is recommended. Analysis should be done by a core lab. There is epidemiological evidence of co-existence of diastolic dysfunction and AF.5 When the diagnostic or therapeutic value of diastolic function is evaluated, transmitral pulsed wave Doppler measurements can be used. M-mode imaging from the parasternal long-axis view gives a unidimensional measurement (antero-posterior direction) and first impression of LA size. Better LA size information can be obtained from 2D or volumetric LA measurements. In patients with sinus rhythm, the maximal A-wave amplitude or its velocity–time integral on pulsed Doppler echocardiography provides information on LA function. Flow velocities (pulsed Doppler imaging) in the LA appendage provide adequate information on LA contractile function, even in AF, but this measurement requires transoesophageal echocardiography.

Highly reliable information on LV volumes and function can be obtained using computed tomography,75 MRI, or gated nuclear imaging.76 In contrast to the real-time imaging obtained by echocardiography, these modalities rely on ECG-based signal averaging. This technical requirement has prevented the use of these techniques in trials of AF patients so far. Furthermore, nuclear imaging techniques and computed tomography require exposure to ionizing radiation.

Heart failure should be a secondary outcome parameter because it is difficult to quantify. NYHA class is widely accepted but not very sensitive to change. Hospitalization for heart failure is a reasonable way to measure the consequences of heart failure in a clinical trial. In addition, VO2(max), 6 min walking distance, and nt-proBNP may be helpful as general measures of cardiac strain and performance.


  • All trials of rate- or rhythm-control interventions should report standard transthoracic echocardiographic data at entry and during follow-up. The assessment should include LA size (M Mode), LV size (M Mode), LV function (M Mode, preferably 2-dimensional echocardiography, modified Simpson technique).

  • For trials assessing other (e.g. antithrombotic) interventions, echocardiography assessment is required at entry and strongly recommended during follow-up.

  • When LV function or heart failure are part of the main outcome parameter set, it is reasonable to supplement echocardiography with a test for exercise capacity (6 min walk test, standardized exercise test) and potentially with a serologic marker (e.g. nt-proBNP).

  • Hospitalizations may serve as an intermediary outcome parameter for heart failure that is easily quantified.

Emerging surrogates as outcome parameters

A limitation of current trials is that AF is considered as one entity. Emerging surrogates will allow to better identify pathophysiological mechanisms underlying AF in a given patient. Different diseases induce different ‘substrates’ for AF. Hence, different forms of AF may require different therapies.7786 In better defined patient populations, it might be easier to demonstrate a therapeutic effect of a given intervention. As an unproven working hypothesis, we propose that the therapy of the cause will be more efficient than the therapy of a symptom. Given the extensive list of potential surrogate markers (Table 8), it is the educated guess of the panel that some of them may develop into novel diagnostic techniques for tiered therapy of AF.10,8790

View this table:
Table 8

Emerging surrogate outcome parameters in atrial fibrillation trials

Surface ECG
 Frequency analysis of fibrillatory activity
 Signal-averaged ECG of the P-wave
 Amplitude of the QRS-complex/markers for left ventricular hypertrophy
 Markers of the autonomic tone (heart rate variability)
 P-on-T ectopic beats
Intracardiac atrial electrograms
 Morphology of atrial electrograms
 Amplitude of atrial electrograms
 Frequency analysis of fibrillatory activity
Blood levels
 Collagen/collagen metabolism (e.g. procollagens, matrix metalloproteinases)
 Inflammatory mediators (e.g. TNF-alpha, interleukins, high-sensitivity C-reactive protein, adhesion molecules)
 Thrombotic markers (e.g. clotting factors, von Willebrandt factor, platelet markers, fibrolytic indices)
 Neurohumoral factors (e.g. angiotensin II, aldosterone, atrial natriuretic peptide, brain-type natriuretic peptide)
 Proteomic profiles
Histological and molecular markers
 Atrial cell size/hypertrophy
 Interstitial fibrosis
 Ultrastructural changes in atrial myocytes
 Components of signalling pathways (e.g. phosphatases, kinases, proteases)

Health economics

The mere number of AF patients and the frequency of complications of the arrhythmia cause large cost, often not visible within traditional health care budgets. Use of new treatment options may reduce this economic burden, but causes additional cost for treatment. It is important that any large-scale controlled AF trial includes a detailed analysis of cost. The result of such exploration should be extended beyond the actual study, projecting the application of the study results on a more general basis. This recommendation is issued with the notice that comparison of health-care related cost is difficult between different health-care systems. Important information related to cost may include hospitalizations including total duration of time spent in a hospital and number and timing of interventions, but also type and duration of medication, time spent on sick-leave, and cost of ambulatory health-care provision.

Specific design issues in atrial fibrillation trials

Composite outcome parameters in atrial fibrillation

Composite outcome parameters should usually be spared for secondary analyses. In any case, the relative contribution of each of these parameters for the composite outcome should be reported and accounted for upfront. At times, a larger or longer trial with less frequent follow-up may yield more important clinical information (e.g. on death or stroke) than a smaller trial with more intensive follow-up details (e.g. on the composite of death, hospitalization, myocardial infarction, and stroke).

Cardiovascular hospitalizations

Cardiovascular hospitalizations have been used as primary outcome parameter or as the main component of a composite primary outcome parameter for AF trials (Table 3). Cardiovascular hospitalizations can be easily and reliably measured in multi-centre trials, but integrate information from several outcome domains, depending on the primary reason for admission, including rhythm, further interventions, heart failure, thromboembolic complications, adverse events, and health economics. Local treatment routines will influence whether a given medical condition is treated on an out-patient basis or in hospital (e.g. cardioversion, initiation of antiarrhythmic drug treatment). Last but not least, a potentially relevant portion of cardiovascular hospitalizations may be unrelated to AF (e.g. myocardial infarctions, see section on AF-related deaths). Hence, number of hospitalizations will measure AF-related and unrelated effects. At times, the time spent in hospital may be more relevant than the number of hospitalizations. The panel recommends to use cardiovascular hospitalizations as a secondary outcome parameter, to report the time (e.g. days) spent in hospital, and to report the contribution of the different causes of cardiovascular hospitalization to this composite outcome parameter.

Further interventions

The necessity for repeated treatment has been used as an outcome parameter in AF trials, especially in trials investigating rhythm-control interventions. Ideally, such ‘further interventions’ are pre-specified parts of the study protocol, rendering them part of the tested (e.g. rhythm control) strategy. The underlying arrhythmia, not the repeated intervention per se, should be defined as outcome parameter.

Time-based assessment of outcome parameters

When a rhythm-based primary outcome event is reached, the patient often ends his/her trial participation. This has been a problem in AF trials. After reaching a time-based (e.g rhythm) primary outcome event, all patients should be followed until the end of the trial to assess death, stroke, and potentially other outcome domains.

‘Blanking’ or ‘therapy stabilization’ periods

Such a period is defined as the time interval during which episodes of recurrent AF should be documented but not counted as components of the ECG-based outcome parameter. As the term ‘blanking period’ can cause confusion among investigators and protocols, it is suggested to use the term ‘stabilization period’. While there are aspects of trial design that strongly argue against such periods, there is sometimes a relevant biological rationale for such a ‘stabilization’ period. If used, however, there are several principles that should be observed: (i) All events during the ‘stabilization period’ need to be recorded and reported. (ii) For reasons of design (intention-to-treat, equal treatment in all study arms), such a ‘stabilization period’ must be pre-specified in the protocol, begin at the time of randomization, and comprise an equal period for all study arms. (iii) All events not related to the ECG-based outcome, e.g. performance measures and adverse events, have to be recorded and counted from the time of randomization.

Assessment of radiation exposure

Catheter ablation procedures and radiation-based imaging (e.g. computed tomography of brain and heart) cause considerable exposure to ionizing radiation for patients91,92 and operators.91,93,94 Assessment of radiation dose is therefore a required part of the safety outcome in every trial that includes fluoroscopy-based interventions or radiation-based imaging, e.g. to guide therapy. Fluoroscopy time is not adequate to measure radiation exposure. Dose–area products (expressed in Gy×cm2) are readily available and recommended to measure radiation energy delivered to the patient and to the operator. They can be used to estimate effective patient-doses.95,96 Radiation exposure must include all exposure including those resulting from pre- or post-procedural imaging (e.g. cardiac CT scans for image fusion).


AF has a complex aetiology and causes morbidity and mortality due to many different mechanisms. A controlled trial in AF patients requires assessment of the effect of therapy in each of the main outcome categories. A careful selection of relevant outcome parameters is mandatory for any AF trial. This paper describes an expert consensus of required outcome parameters in seven relevant outcome domains and gives information on additional, more intensive monitoring of outcome. Although this exceeds current practice in some trials, the panel recommends basal assessment of all major outcome domains in AF trials. Only such a comprehensive set of outcome parameters will allow to compare the effects of different treatments across trials.


The Consensus Conference and the preparation of the two consensus documents simultaneously published in the European Heart Journal (Executive Summary) and in Europace (Full Paper) have been funded by the European Heart Rhythm Association (EHRA) and by the German Atrial Fibrillation competence NETwork (AFNET). AFNET receives funds from the Federal German Ministry of Education and Research (BMBF, 01Gi0204). In addition, participation fees were used to cover expenses.


We would like to thank the staff of the Heart House, Sophia Antipolis, especially Sophie Janiszewski, as well as the AFNET staff, especially Angelika Leute, for excellent support of the conference.

Conflict of interest: Many of the authors of this document have been involved in the design, conduct, and analysis of clinical trials in AF, and have been supported by industry or other research financing bodies.

AFNET and EHRA support a policy of transparency towards such involvements as required by the ESC journals. Therefore, the following table lists potential conflicts of interest of the authors. Furthermore, corporate affiliations of the other contributors to this document, who are listed in the Appendix section are indicated after the respective names.

NameConsulting fees/honorariaSpeaker's bureauOwnership/partnership/employeeResearch grantsPrincipal investigator (PI)/Steering committee (SC)Salary
Paulus Kirchhof3M Medica
NoneNoneCardiovascular Therapeutics
3M Medica/MEDAPharma
German Federal Ministry
for Education and
Research (BMBF)
Fondation LeDucq
German Resarch
Foundation (DFG)
Angelo AuricchioMedtronic
Boston Scientific
Boston Scientific
Jeroen BaxNoneNoneNoneBoston/Guidant
GE Medical
St. Jude Medical
Harry CrijnsSanofi-Aventis
St. Jude
St. Jude Medical
John CammSanofi-Aventis
Bristol Myer
Bristol Myer Squibb
Hans-Christoph DienerAbbott
Bayer Vital
Boehringer Ingelheim
Janssen Cilag
Boehringer Ingelheim
Bertelsmann Foundation
Heinz-Nixdorf Foundation
European Union
German Federal Ministry
for Education and
German Resarch
Foundation (DFG)
Andreas GoetteDaiichi-Sankyo
3M Medica
NoneNoneGerman Federal Ministry for
Education and Research
3M Medica/MEDAPharma
Gerd HindricksBard
St. Jude Medical
St. Jude
NoneSt. Jude Medical
Stefan HohnloserSt. Jude Medical
Boehringer Ingelheim
Sanofi-AventisNoneSt. Jude Medical
Boehringer Ingelheim
Numerous trials in
the past and
(sponsors to
the right)
Lukas KappenbergerMedtronicNoneNoneMedtronic
Karl-Heinz KuckBiosense-Webster
St. Jude Medical
St. Jude Medical
Gregory Y.H. LipAstraZeneca
Daiichi Sankyo
Medical Research Council
Trials sponsored by AstraZeneca
Bertil OlssonAstraZeneca
Swedish Heart Lung
Lund Univ/Hosp
Local foundations
Stroke prevention studies with AZNone
Thomas Meinertz3M Medica
German Federal Ministry
for Education and
Research (BMBF)
Silvia PrioriPfizer
Reliant Pharmaceuticals
Medifacts Corp.
Ursula RavensCardiome Pharma Corporation
NoneNoneGerman Federal Ministry for Education and Research (BMBF)
Fondation Leducq
German Resarch Foundation (DFG)
Gerhard SteinbeckMedtronic
St. Jude Medical
Astra Zeneca
St. Jude Medical
Elisabeth SvernhageNoneNoneEmployee of AstraZenecaNoneNoneYes
Jan TijssenNoneNoneNoneNoneNoneNone
Alphons VincentNoneNoneEmployee of MedtronicNoneNoneYes
Günter BreithardtBoston Scientific
NoneNone3M Medica/MEDAPharma
CAD-Ref Registry


Additional contributors: A.M. Baczynska (EMEA, London, Great Britain); S. Berkowitz (Bayer. Leverkusen, Germany); R. Bilke (Boehringer, Ingelheim, Germany); C. Blomstrom-Lundqvist (Uppsala, Sweden); A. Bollmann (Magdeburg, Germany); M. Brignole (Lavagna, Italy); J. Brugada (Spain); N. Edvardsson (AstraZeneca, Sweden); T. Fetsch (IKKF, München, Germany); K.G. Häusler (Berlin, Germany); W. Haverkamp (Berlin, Germany); H. Heidbüchel (Leuven, Belgium); E. Hoffmann (Munich, Germany); A. Huemmer (St. Jude Medical, Frankfurt, Germany); C. Israel (Frankfurt, Germany); E. Köfüncü (Sanofi-Aventis. Berlin, Germany); K.-H. Ladwig (Munich, Germany); S. Lévy (Marseille, France); F. Lindemans (Medtronic, Maastricht, the Netherlands); J. Merino (Madrid, Spain); F. Misselwitz (Bayer); M. Näbauer (Munich, Germany); M. O' Donnell (Biosense Webster); M. Oeff (Brandenburg, Germany); N. Osypka-Rubenstein (Osypka, Grenzach-Whylen, Germany); R. Peeters (Sorin); P. Ramge (Sanofi-Aventis, Berlin, Germany); S. Schepels (Biosense Webster); M. Schwertfeger (Sanofi-Aventis, Berlin, Germany); C. Stoeppler (St. Jude Medical, Frankfurt, Germany); R. Sutton (London, Great Britain); I. Van Gelder (Groningen, the Netherlands); P. Vardas (Crete, Greece); T. Weiß (AFNET, Münster, Germany); S. Willems (Hamburg, Germany); R. Woker (Boehringer, Ingelheim, Germany).


View Abstract