OUP user menu

Paroxysmal atrial fibrillation

G.Y.H. Lip, F.L. Li Saw Hee
DOI: http://dx.doi.org/10.1093/qjmed/94.12.665 665-678 First published online: 1 December 2001


There has been a tendency to treat paroxysmal atrial fibrillation (PAF) in a similar way to sustained AF, but treatment objectives may be very different. We discuss current definitions, epidemiology, pathophysiology and natural history of PAF, and review evidence for its treatment and management. PAF comprises between 25% and 62% of cases of AF, with similar underlying causes to those in sustained AF. The main objective of management is prevention of paroxysms and long‐term maintenance of sinus rhythm, and Class 1c drugs are highly effective, although beta‐blockers are useful alternatives. If patients have severe coronary artery disease or poor ventricular function, amiodarone is probably the drug of choice. Although randomized controlled trials of thromboprophylaxis in patients with paroxysmal AF per se are lacking, the approach to patients with paroxysmal AF should be similar to that in patients with sustained AF, with warfarin for ‘high risk’ patients and aspirin for those at ‘low risk’. Non‐pharmacological therapeutic options, including pacemakers, electrophysiological techniques and the implantable atrial defibrillator, show great promise. Despite paroxysmal AF being a common condition, management strategies are limited by evidence from small randomized trials, with inconsistencies over the definition of the arrhythmia and the inclusion of only symptomatic subjects. Evidence for antithrombotic therapy is also based on epidemiological studies and subgroup analyses of the large randomized trials.


Atrial fibrillation (AF) is the commonest sustained disorder of cardiac rhythm, which is often associated with a high risk of morbidity and mortality from heart failure, stroke and thromboembolic complications. There have been significant advances in our understanding of AF, but much of our knowledge of the epidemiology, clinical presentation and management strategies of this arrhythmia have been based on studies of patients predominantly with chronic (sustained) AF.

Nevertheless, AF may also occur intermittently, and the importance of paroxysmal AF (PAF) has recently gained prominence. A common error in clinical management is to treat chronic sustained AF and paroxysmal AF similarly, despite some differences in management objectives. PAF may be associated with risks of stroke and thromboembolism similar to those for sustained AF, and many patients suffer significant morbidity. Recent advances in areas of electrophysiology and pathophysiology of AF have also rekindled much interest in PAF.

The objective of this review is to discuss current definitions of AF, the epidemiology, pathophysiology and natural history of paroxysmal AF, and importantly, to review the published evidence for the treatment and management of paroxysmal AF.

Review methods

Data Sources

We searched the Cochrane Library and the US National Library of Medicine's Medline for up‐to‐date reviews and randomized controlled trials (RCTs) of the treatment and management of PAF. If no RCTs had been done, we relied on observational cohort and case‐controlled studies, especially for information on the epidemiology, pathophysiology and natural history of PAF. We did not undertake a detailed systematic review of aspects of PAF management for which there have been only limited or no reliable studies with large patient groups (e.g. the implantable atrial defibrillator (atrioverter) and surgery).

Study selection and data extraction

We included studies with clearly defined populations with PAF and treatment outcomes, which were selected and extracted by a consensus of two reviewers.

Definitions of atrial fibrillation

There are many definitions of AF, but one proposed definition divides patients into acute‐onset and chronic AF1 (Figure 1). The latter are further divided into paroxysmal and the more sustained forms, that is, persistent and permanent AF. In paroxysmal AF, the main objective of management is the prevention of paroxysms and long‐term maintenance of sinus rhythm, as well as appropriate consideration of antithrombotic therapy. In persistent AF, the possibility of cardioversion from AF to sinus rhythm remains, and the objective of management is appropriate anticoagulation to reduce the risk of thromboembolism with cardioversion, and anti‐arrhythmic therapy to maintain long‐term sinus rhythm post‐cardioversion. In permanent AF, cardioversion is not feasible or has failed, and the objective of management is heart rate control and appropriate antithrombotic therapy.

These subcategories may coexist in the same patient, and objective data to support a rigid classification into the three categories are still lacking; indeed, the practical management principles are still broadly similar between the subgroups, with attention to preventing thromboembolism and the appropriate approach to dealing with the arrhythmia (rhythm control or rate control).1

Defining paroxysmal atrial fibrillation

The study of PAF in the past has been bedevilled by inconsistencies over its definition and the inclusion of various forms of AF and patients with paroxysmal supraventricular tachycardia in many small series. Furthermore, only subjects with symptomatic PAF were included in some of these studies. The precise definition of PAF is also difficult, as reported series vary considerably in defining the length and duration of episodes.

Thus, the paroxysmal form of AF comprises a heterogeneous group of patients that may differ in the frequency, duration, mode of termination and the presence or severity of symptoms. However, the differentiation of PAF from sustained (that is, persistent or permanent) AF has often been based on the history given by the patient, provided he/she is symptomatic. This can be misleading, as in one study asymptomatic PAF occurred 12 times more commonly than symptomatic PAF in patients when followed up longitudinally by Holter monitoring.2

PAF is usually defined temporally, as intermittent periods of AF interspersed with episodes of normal sinus rhythm, normally lasting <7 days,3 although some episodes of PAF will last longer. PAF may be self‐terminating in <48 h or be more persistent, lasting >48 h (the arbitrary time after which the patient is deemed to require formal anticoagulation prior to cardioversion). However, in the same patients, the characteristics of arrhythmia presentation may change with time.1,,3

Figure 1

A clinical classification of atrial fibrillation and the approach to management.

Epidemiology of paroxysmal atrial fibrillation

Estimates of the prevalence of AF vary widely around the world. In the Framingham study, the prevalence of AF ranged from 0.5% at age 50–59 years to 8.8% at 80–89 years.4 Sudlow and colleagues5 estimated that AF may in fact be more prevalent in UK than in the US or Australia, and suggested that between 160000 and 644000 patients aged >65 years in Great Britain might have AF. However most of these estimates would apply to permanent AF, and the data for PAF is more limited.

PAF is estimated to comprise between 25% and 62% of cases of AF seen by practitioners in both hospital and primary care setting.6–,10 The reported prevalence varies widely, because of differences in definitions and the various populations studied. In addition, the prevalence of PAF could be underestimated, as most epidemiological studies depend on symptomatic episodes, but asymptomatic PAF is common on Holter monitoring.2

For example, Takahashi and colleagues8 described a consecutive hospital‐based series of 234 patients with AF, and found that 94 (40.2%) had a paroxysmal form of AF. Of the 1212 patients with AF admitted to hospital in Denmark, Godfredsen and colleagues6,,7 found that 426 (35%) had PAF, and 786 (65%) had sustained AF. The Stroke Prevention in Atrial Fibrillation (SPAF) investigators11 defined PAF as the presence of sinus rhythm at either the 1‐ or 3‐month follow‐up visit after enrollment, and found that 486 (27.8%) of recruited subjects had intermittent AF and 1263 subjects had chronic sustained AF.

The prevalence of PAF has also been estimated in population‐based studies. For example, the Framingham Study suggested a presence of 25%, but their definition was based on AF not being present on routine biennial examination after being present at a previous hospitalization.9 In contrast, a population study from Minnesota found that 62% of all AF cases were intermittent.10

The underlying causes of PAF and sustained AF are similar (hypertension, coronary artery disease, rheumatic heart disease, etc.), although there appears to be great variability amongst the studies. No underlying cause can be found in nearly half of PAF patients.12,,13 By contrast, sustained ‘lone’ AF has a prevalence of only 0.7%.7 Compared to patients with chronic sustained AF, patients with PAF are usually younger, have less hypertension and congestive heart failure, and by echocardiography have smaller left atria and better left ventricular systolic function.11

Natural history of paroxysmal atrial fibrillation

The natural history of PAF is unclear, because of the difficulty in defining the patient population precisely and obtaining its actual incidence. This has led to some difficulties in defining the true natural history of PAF, but many cases eventually become sustained AF if not treated. In patients with PAF, the recurrence of AF differs in different reports, varying from 70% at one year in one study (without anti‐arrhythmic treatment)14 to 90% at 4 years.15

One of the more recent prospective series of patients with PAF was reported by the Stroke Prevention in Atrial Fibrillation (SPAF) group,11 which found a recurrence rate of 49% over a period of 26 months. Clinical predictors for recurrence of AF derived from SPAF data11 included older age group, prior myocardial infarction and the presence of congestive cardiac failure. Echocardiographic predictors included an enlarged left atrium, significant left ventricular wall motion abnormalities and a lower left atrial appendage flow velocity of <30 cm/s on transoesophageal echocardiography and the presence of significant mitral regurgitation.16

Clinical experience nevertheless suggests that AF is a self‐promoting electrical disease, with PAF frequently progressing to sustained AF.12,,17 Indeed, sinus rhythm becomes less sustainable after electrical or pharmacological conversion of AF, the longer AF lasts.18,,19 It has been estimated that about 25% of PAF cases eventually become persistent or permanent, with the majority of the latter having underlying heart disease.20 Takahashi and colleagues8 found that 25% of their PAF patients converted to chronic sustained AF over a period of 10 years. By contrast, the Danish group6,,7 found that 33% of their PAF patients progressed to sustained AF after a median follow‐up of 9 years (range 0–24 years). The rate of progression to permanent AF varied according to aetiology, with patients with rheumatic valve disease having the highest rate of progression (66%) when compared to a rate of 40% in hypertension and 27% in patients with coronary heart disease.

The development of permanent AF may also be related to the duration of paroxysms of AF. When PAF lasts <2 days, conversion to permanent AF occurs in 31% of patients, but if PAF lasts >2 days, permanent AF occurs in 46%. These clinical observations are in keeping with an experimental model using conscious goats, in that the more frequently AF is induced or the longer the AF is maintained, the more easily inducible and the more permanent AF becomes and the more difficult it is to revert to sinus rhythm (‘AF begets AF’).21


The main pathophysiological implications of AF are related to loss of atrial transport, the irregular fast ventricular response and increased risk of thromboembolism. In a patient with PAF, the change in rhythm from sinus rhythm to AF results in loss of atrial systole, leading to haemodynamic effects and increased atrial stasis. The latter contributes to thrombogenesis and the risk of stroke.

In older patients, especially in those with associated hypertension or diastolic dysfunction, atrial systole may account for up to a third of stroke volume; thus the loss of atrial transport leads to the haemodynamic manifestations of reduced exercise tolerance, exertional dyspnoea, tiredness and fatigue.1 The irregular and fast ventricular response results in reduced cardiac efficacy and if the ventricular response is uncontrolled in the long term, this can lead to progressive ventricular dilatation and impaired left ventricular systolic function, often referred to as a ‘tachycardia‐induced cardiomyopathy’.22 The fibrillating atrium can also lead to atrial stasis and abnormalities of haemostasis, platelets and endothelial dysfunction, which confers a hypercoagulable state, present in both permanent and paroxysmal AF.23,,24


Several potential electrophysiologic mechanisms may be responsible for paroxysmal AF. In general, two theories have been suggested as the underlying mechanism for AF, namely:25 (i) enhanced automaticity, involving one or more foci firing rapidly; or (ii) re‐entry involving one or more circuits. In addition, heterogeneity of atrial refractoriness and slow atrial conduction times (allowing time for the myocardium to regain excitability between each wavefront) lead to AF being sustained in the long term. In the presence of sinoatrial disease, these manifest as the tachy‐brady syndrome, with paroxysmal atrial arrhythmias (including paroxysmal AF) alternating with episodes of sinus bradycardias. Implantation of an atrial pacemaker results in stabilization of the atria electrically and suppression of paroxysmal AF.25

In selected patients, paroxysmal AF may be the result of either focal sources or stable rentrant circuits that drive the remaining atrial tissue until degeneration to AF occurs. Jais et al.26 first reported nine patients with drug‐resistant paroxysmal AF, in whom a single, rapidly firing focus was identified using electrophysiological mapping, with a centrifugal and consistent pattern of atrial activation and striking and abrupt changes in atrial cycle lengths. The ablation of these foci, near the ostia of great vessels, resulted in a cure in these relatively young patients. The pulmonary veins are an important source of ectopic beats, initiating frequent paroxysms of AF, and that these foci also respond well to treatment with radio‐frequency ablation.27

The autonomic system and the substrate for paroxysmal AF

There is some evidence, from laboratory studies of cellular electrophysiology, studies in isolated hearts and intact animals and humans, through to clinical observations in patients, on the influence of the autonomic nervous system on the recurrence of AF.28 A role for the autonomic nervous system in PAF was also demonstrated by Yamashita and colleagues,29 who showed that PAF exhibited a unique circadian variation in total duration of AF that differed from the well‐known pattern for acute cardiovascular events. Although there was no relationship between time of day and the onset of AF, there was a circadian rhythm of the total duration of AF, with the longest duration of AF when its onset was during the night, peaking at 12 am, whilst the shortest duration occurred when the onset was at 11 am.29

Various attempts at delineating the substrate for AF and interplay between sympathetic or vagal activities in initiating PAF have used P‐wave signal‐averaged electrocardiograms and heart rate variability. A prolonged signal‐averaged P‐wave duration and the presence of low‐voltage atrial late potentials may be useful in identifying patients at risk for developing PAF,30–,32 including those who have undergone ablation of an accessory bypass tract.33 Indeed, P‐wave‐triggered signal‐averaged ECG may have some prognostic value, as a P wave duration ≥145 ms and root mean square voltage for the last 30 ms <3.0 mV can predict the development of permanent AF. A recent prospective study of 75 patients found that an abnormal P‐wave signal‐averaged ECG (hazard ratio 19.1, p=0.0069) and high atrial natriuretic peptide levels (hazard ratio 8.6, p=0.018) were independent predictors of PAF development in patients with left ventricular ejection fraction <40%, over the 21‐month follow‐up period.34

However, the role of these techniques in predicting the occurrence of AF in other patient groups, and in diagnostic assessment and therapeutic decision‐making in patients with PAF needs to be clarified by controlled studies.35 The onset of AF at night may be preceded by an increase in high‐frequency components of heart rate variability, which does not appear to be the case for episodes that occur during the day. The heart rate in sympathetically‐mediated AF is usually higher before and during the episode, compared to the vagal type of AF.

Both vagal and adrenergic forms of PAF36 have been characterized, based upon differences in clinical presentation. The ‘vagal form’ typically has a predominance of men, age of onset at 40 to 50 years and a higher prevalence of lone AF. Vagal AF is usually preceded by bradycardia, and tends to occur at night, during rest, after eating, or with absorption of alcohol. Importantly, however, both beta‐blockers and digoxin may increase the frequency of the vagal form of AF. By contrast, the ‘adrenergic’ form of AF form is less common than the vagal type, and tends to occur only during daytime, being preceded by exercise, adrenergic stimuli (such as exercise) and emotional stress. Thus beta‐blockers are usually the treatment of choice in adrenergic AF. However, manifestations of both vagal and adrenergic AF may be present in the same patient, and the clinical pattern may differ over the course of time.36

Auto‐immune process

Recently, Maixent and colleagues37 demonstrated the presence of circulating auto‐antibodies against myosin heavy chain in a significant proportion of PAF patients. This raises the possibility of an auto‐immune process in some patients with PAF, but this may be related to provision of an anatomic substrate in the atria of these patients, precipitating AF.

Thromboembolic risk

Clinical evidence of thromboembolism in paroxysmal atrial fibrillation

Early data suggested that the risk of stroke was greater at the onset or offset of PAF and the patients with PAF may be at an increased risk of stroke compared to those with permanent AF.20 However, subsequent studies reported that the risk of thromboembolic complications for PAF may be similar or even less than that of permanent AF.

The Framingham Study38 found a low annual incidence of stroke for patients with PAF of about 1.3%. By contrast, Peterson and colleagues6 reported an annual incidence of stroke in patients with PAF of 2.0%, that increased to 5.6% with permanent AF. In contrast, the SPAF study39 found no significant difference in stroke risk with PAF compared to those with permanent AF, a finding also noted in the Boston Area Trial of Atrial Fibrillation Investigators Study.40 Indeed, the pooled analysis of the AF Investigators concluded that patients with PAF had a similar stroke risk to patients with permanent AF, which could be reduced by antithrombotic therapy. Further, the type of AF (PAF or permanent) and the length of time the patient was in AF had no effect on the stroke rate.41 A histopathological case‐control study by Yamanouchi and colleagues42 of brains from 54 consecutive PAF patients aged 70 years or older found that cerebral infarctions were present in 54% of PAF patients, compared to only 22% of controls.

This variation observed in risks of stroke and thromboembolism in PAF may well be related to the different study populations, and possibly the frequency and duration of the PAF.23 Furthermore, data based on the randomized trials should be interpreted with caution, as only a small number of patients screened were actively entered into the trials. The range in thromboembolic risk is likely to be wide, as the risk in a patient with PAF with a single short paroxysm once a year is likely to be different from that in a patient with long, daily paroxysms.23 A further contribution to stroke risk is likely to be age and the presence of underlying structural heart disease.

Mechanical factors

On theoretical grounds, the change from the irregular, uncoordinated atrial activity of AF to the regular, synchronous atrial contractions in sinus rhythm (and vice versa) could potentially result in embolization of pre‐existing thrombus in patients with PAF. This may partly explain why there is often a cluster of stroke and thromboembolism when PAF becomes more persistent or permanent, and the high incidence of thromboembolism in the immediate post‐cardioversion period, in the absence of anticoagulation. The intermittent nature of atrial stasis is elegantly demonstrated by the presence of spontaneous echo‐contrast on transoesophageal echocardiography, indicating atrial stasis, during paroxysms of AF.43

Haemostatic abnormalities

Patients with PAF demonstrate abnormalities of haemostasis consistent with a hypercoagulable state44–,47 that also appear to be related to the duration of paroxysms of AF and whether the patient was in AF at the time of sampling.46

In the study by Sohara et al.,46 markers of platelet activation (such as beta‐thromboglobulin and platelet factor 4) and fibrinogen were elevated during episodes of AF that were >12 h in duration. In the study by Lip et al.,44 patients with PAF had intermediate levels of plasma fibrinogen and fibrin D‐dimer (an index of thrombogenesis) when compared to levels in patients with chronic AF and controls in sinus rhythm. Patients who are started on anticoagulation48 or cardioverted from AF to sinus rhythm45,,49 demonstrate significant changes in markers of haemostasis, suggesting a reduction in the hypercoagulable state. Nevertheless, the study by Li Saw Hee et al.47 suggests that there may be differences in the prothrombotic state between patients with permanent, paroxysmal and persistent AF, with abnormal haemostasis in permanent and paroxysmal AF, while indices of hypercoagulability were not different between persistent AF and controls in sinus rhythm. Thus, interpretation of studies of the hypercoagulable state in AF may be dependent upon the definition of the patient group studied.

The clustering of thromboembolic events during the transition from PAF to permanent AF could be explained by the transient haemodynamic and haemostatic abnormalities in PAF.23 Currently, there are no data regarding the value of elevated haemostatic markers for predicting future stroke in patients with permanent AF or PAF.

Management of paroxysmal atrial fibrillation

The three main aims of treatment for PAF are: (i) to suppress paroxysms of AF and maintain long‐term sinus rhythm; (ii) to control heart rate during paroxysms of AF if they occur; and (iii) to prevent the complications associated with PAF, i.e. stroke and tachycardia‐induced cardiomyopathy.1 Many patients with PAF are often highly symptomatic, and the psychological aspects of their management should not be neglected. Paroxysms of AF which are persistent (that is, lasting >48 h) should be considered for cardioversion to sinus rhythm.

The management of PAF patients should include the recognition that abolition of symptoms of PAF does not necessarily mean abolition of the arrhythmia, as heart rate slowing may abolish symptoms, and asymptomatic episodes are common.50 It may be appropriate to document the frequency of arrhythmia by Holter monitoring in patients with symptomatic recurrences, and to repeat monitoring if symptoms are abolished by therapy, to ascertain whether asymptomatic episodes of AF are present.1 Nevertheless, it should be noted that evidence for management strategies in paroxysmal AF is limited by evidence from relatively small RCTs of short duration, which have been performed in (mainly) symptomatic subjects. Even the evidence for antithrombotic therapy in PAF is also based on extrapolation from epidemiological studies and subgroup analyses of large RCTs.

Prevention of paroxysms and maintenance of sinus rhythm

If attacks of PAF are frequent, chronic prophylaxis with drugs can be effective, after removal of precipitating factors such as caffeine, alcohol, stress, and adequate treatment of underlying diseases such as myocardial ischaemia, thyrotoxicosis, and heart failure. Nevertheless, the prevention of PAF with medication has been plagued by high discontinuation rates, due to side‐effects and drug inefficacy. Another important limitation of the published evidence is that many pharmacological studies of PAF have concentrated on the reduction of symptomatic recurrences of PAF, and as discussed earlier, many patients with PAF have frequent asymptomatic episodes. Many agents also have proarrhythmic tendencies and can increase the risk of sudden death, especially if poor left ventricular function is present. In addition, many studies in PAF have often involved mixed populations, including both PAF, paroxysmal atrial flutter and persistent AF. Finally, many older drugs, such as procainamide, have not been tested extensively in appropriate placebo‐controlled studies of PAF.

In the long term, very few patients achieve complete suppression of paroxysms of AF, and even the best pharmacological therapy to maintain sinus rhythm succeeds in only 70% of patients.51,,52 One obvious implication is that it cannot be assumed that patients can be rendered absolutely free of any risk of thromboembolism by anti‐arrhythmic therapy. The AFFIRM study53 is currently examining whether the strategy of aggressive maintenance of sinus rhythm (‘rhythm control‘) will improve clinical endpoints over a strategy of heart rate control and anticoagulation (‘rate control‘), and has yet to report its findings. Furthermore, pharmacological and electrical cardioversion carries a similar thromboembolic risk, which may possibly be minimised by transoesophageal echocardiography, which can be used to guide cardioversion by optimising use of thromboprophylaxis.54

Class 1c drugs are highly effective in PAF, but should not be given to individuals with significant structural cardiac disease, including those patients with a previous myocardial infarction or poor left ventricular function.55,,56 However, in patients with structurally normal hearts, Class 1c drugs, such as flecainide or propafenone, would be the drugs of choice. The incidence of proarrhythmia is <3% for flecainide,57,,58 and <2% for propafenone,59,,60 which can be minimized if electrolyte abnormalities and drugs that prolong QT interval are avoided.

When used orally to prevent AF, some 60–70% of patients taking flecainide remain in sinus rhythm at 1 year. Flecainide has the same frequency of side‐effects as propafenone, but is better tolerated than quinidine, being discontinued in 5–10% because of side‐effects.61,,62 For example, Nacarelli et al.61 reported that both flecainide and quinidine were equally effective in adequately controlling recurrences of PAF, but fewer flecainide patients discontinued therapy due to side‐effects. Propafenone also has a mild beta‐blocking activity and when used orally for prevention of AF recurrence; 50–60% of patients remain in sinus rhythm at 1 year.62 One recent randomized placebo‐controlled trial confirmed the efficacy of propafenone in the prophylaxis of paroxysmal supraventricular tachycardia and paroxysmal AF, with few adverse effects.60 Propafenone is at least as effective as sotalol and flecainide, and perhaps more effective than quinidine.

In patients with PAF, especially those of the adrenergic form, a beta‐blocker may be the drug of choice.1 Some prefer agents such as sotalol, which has both Class III and beta‐blocking activity, but at the doses used in PAF (usually ≤80 mg t.i.d.), the predominant pharmacological effect is from beta‐blockade.1 This has led to less use of sotalol, especially since this drug can be associated with proarrhythmia, such as torsades des pointes, although this risk is low if the dose is kept below 80 mg b.d. One recent open‐label, randomized, crossover study comparing sotalol 80 mg b.d. and atenolol 50 mg daily found no difference in ECG or symptomatic control of PAF between the two treatment groups.63

Amiodarone has been increasingly used in patients with recurrent AF, especially if patients have heart failure or poor left ventricular function on echocardiography. Low doses are also generally well‐tolerated, and may suffice for treatment of paroxysmal AF, with minimal risk of side‐effects. A combined analysis of available trials reported that amiodarone was effective in 60–70% of patients with drug‐refractory AF.51,,64 However, there has been a lack of prospective data on the role of amiodarone as the first‐line treatment of patients with recurrence of AF, until the recent Canadian Trial of Atrial Fibrillation (CTAF) study, which compared amiodarone with sotalol and propafenone in 403 patients.65 After 16 months, 63% of patients taking sotalol or propafenone had a recurrence of AF, compared with 35% of those taking amiodarone.

Newer class III agents, such as ibutilide, dofetilide and azimilide, have been found to be effective in cardioversion of (persistent) AF, but ongoing studies are still examining the value of these agents in preventing PAF. Results from the DIAMOND study (Danish Investigators of Arrhythmia and Mortality On Dofetilide) study,66 which randomized 1518 patients with symptomatic heart failure and left ventricular dysfunction to receive either dofetilide or placebo are encouraging—in a double‐blind clinical trial with a median follow‐up of 18 months, where AF was present in 190 cases in the dofetilide group and 201 in the placebo group, dofetilide cardioverted 12% of heart failure patients in AF to sinus rhythm after one month (vs. 1% with placebo). In addition, dofetilide was significantly better at maintaining sinus rhythm, and reducing the risk of hospitalization for heart failure. However, dofetilide was associated with torsade de pointes in 3.3% of patients, compared with no cases in the placebo group. Although there was no mortality difference between the two groups, treatment was initiated in hospital with 3 days of intensive cardiac monitoring.

Digoxin should probably be avoided in PAF. Paroxysms of AF occur much more frequently and for significantly longer in patients receiving digoxin.67,,68 Further, during an episode of paroxysmal AF, the initial heart rate is similar with or without digoxin.68 The mechanism for this is unclear, but digoxin increases vagal tone, moderating the speed of atrio‐ventricular conduction, and also reduces the atrial refractory period. This latter property may paradoxically render the atrium more susceptible to AF and may reduce or even prevent the chance of reversion to sinus rhythm.

In the CRAFT study,69 where 43 patients with frequent symptomatic paroxysmal AF were randomized to a double‐blind crossover comparison of digoxin and placebo, digoxin reduced the frequency of symptomatic AF episodes and increased the median time between episodes. However, the effect was small and may be due to a reduction in the ventricular rate or irregularity rather than a direct anti‐arrhythmic action. Digoxin is also ineffective for the long‐term maintenance of sinus rhythm or for the cardioversion of AF.70

Adrenergic and vagal forms of PAF

Unequivocal identification of vagal or adrenergic AF may be almost impossible, and the division may sometimes be artificial, except in a small proportion of extreme cases. Nevertheless, in patients suffering from predominantly vagally‐induced AF, beta‐blockers and digitalis should be avoided as these drugs may provoke attacks.36 By contrast, quinidine, disopyramide and flecainide may be effective due to their vagolytic properties. Propafenone is also considered ineffective due to its beta‐blocking properties.

In patients with predominantly adrenergic‐dependent AF, however, underlying cardiac disorders should be treated. After that, patients usually benefit from a beta‐blocker.36 Class 1A and 1C drugs are usually ineffective, although some patients may respond to propafenone. However, digitalis, beta‐blockers or the rate‐limiting calcium‐blockers (verapamil, diltiazem) may be necessary to control the ventricular rate when a relapse of AF occurs, which may be the case in a third of PAF patients despite anti‐arrhythmic therapy. Such agents are also needed to prevent rate‐dependent proarrhythmias of Class 1A and 1C drugs during a recurrence of AF.71 Indeed, the use of anti‐arrhythmic agents such as quinidine and disopyramide with anticholinergic properties may change atrial flutter into 1:1 conduction unless a rate‐limiting agent is also used.72 Controlling the ventricular rate in patients with PAF in the setting of sick sinus syndrome may occasionally be impossible without implanting an artificial pacemaker.

Cardioversion of persisting episodes of atrial fibrillation

In some patients, paroxysms of AF last longer than 48 h, and consideration of cardioversion to sinus rhythm is needed. Early cardioversion of persisting AF would avoid electrical remodelling and structural changes in the atria that result in persistence of the arrhythmia and reduction in the probability of successful reversion to sinus rhythm (‘AF begets AF’).21

Cardioversion can be performed both electrically and pharmacologically, the latter using Class I and III agents. Conversion rates of up to 90% are found 1 h after intravenous flecainide or propafenone.73–,76 Both flecainide and propafenone can also be administered orally, with success rates >70% at 8 h.76 Class III anti‐arrhythmic drugs perform less well, especially in terms of acute (<1 h) conversion.

Patients admitted with new‐onset AF who are admitted should be started on intravenous heparin, and if it can be reliably determined that AF onset was <48 h, cardioversion can be performed without the need for long‐term anticoagulation if this was the first, single paroxysm of AF.1 However, if the onset of AF is uncertain, patients need anticoagulation, especially in the post‐cardioversion period, to reduce thromboembolism.1

The other treatment option of PAF is the ‘pill‐in‐the‐pocket’ approach where patients not on regular therapy take one or two doses of an anti‐arrhythmic agent only when they experience AF, hoping to shorten the duration of the episodes.76 For example, a single oral loading dose of propafenone 600 mg or flecainide 300 mg has reported success rates of 50% at 3 h and 80% at 8 h for both drugs.76 This may be effective in some patients. However this method can only be implemented after the drug has been used under supervision and documented not to produce adverse effects, and there is good and sensible patient compliance with the prescribed dose. Such empirical therapy may be dangerous if taken beyond 48 h, in inadequately anticoagulated patients, or if taken too frequently or infrequently.

Antithrombotic therapy

Nonvalvular AF confers a substantial risk for stroke and thromboembolism, which is estimated to be between 4.5% and 12% per year, depending on associated risk factors.41 Recent studies have established the value of warfarin as thromboprophylaxis in AF; but this treatment carries with it the inconvenience of regular monitoring of anticoagulation intensity and the risk of bleeding. However, none of the trials specifically investigated thromboprophylaxis in PAF per se. In the analysis by Hart et al.77 of patients with intermittent AF in the Stroke Prevention in AF study, they were generally younger (66 vs. 70 years, p<0.001), more often women (37% vs. 26%, p<0.001) and less often had heart failure (11% vs. 21%, p<0.001) than those with permanent AF. Importantly, however, the annual rate of ischaemic stroke was similar for those with intermittent (3.2%) and sustained AF (3.3%). As both sustained AF and PAF confer a similar thromboembolic risk especially in the presence of risk factors (such as age, hypertension, prior stroke, etc.), the approach to patients with PAF should essentially be similar to the subjects with permanent AF.

Risk stratification is therefore important in the management of both permanent AF and PAF; most thromboembolic risk stratification can be performed on clinical criteria alone, with some refinement using echocardiography.78 Typical ‘high risk’ features include previous stroke or transient ischaemic attack, hypertension, congestive heart failure, diabetes and other vascular disease, whilst echocardiography adds further to this if moderate‐severe left ventricular systolic dysfunction is seen on two‐dimensional echocardiography.79 Indeed, in the pooled echocardiographic analysis from the AF Investigators of 1066 patients from three clinical trials, isolated left atrial dilatation on M‐mode echocardiography per se was no longer an independent risk factor for stroke on multivariate analysis.79 In general, our recommendation is to use warfarin in patients with PAF unless they can be classified as having ‘lone’ AF or there are contraindications to the use of anticoagulation, where aspirin should be used instead.

Non‐pharmacological treatment of paroxysmal atrial fibrillation

Episodes of persistent AF have long been effectively treated with external, transthoracic defibrillation. Despite concomitant, post‐cardioversion therapy with anti‐arrhythmic agents, patients will often have additional episodes of AF (up to 50% to 60%) requiring either further repeat external cardioversion or treatment with either pharmacological or additional non‐pharmacological therapies. The limited long‐term efficacy of different treatment regimens has resulted in the development and evaluation of newer, non‐pharmacological therapeutic options for PAF, including electrophysiological techniques and the implantable atrial defibrillator.

Pacemaker therapy

In patients with sinus node dysfunction and PAF, the choice of particular pacemakers may prevent the development of AF in some circumstances, and certain modes of pacing are better suited to patients who have episodic or permanent AF. For example, the randomized prospective study by Anderson et al.80 showed a reduced frequency of thromboembolic events and episodes of AF with atrial pacing when compared with ventricular pacing. More importantly, further follow‐up analysis showed an additive reduction in mortality and heart failure, with a slow progression to complete atrioventricular block in the absence of pre‐existing conduction abnormalities.81 Other trials such as the Canadian Trial of Physiological Pacing (CTOPP) are comparing VVIR pacing with physiological pacing (AAIR or DDD) in patients with either sinus node disease or AV conduction disease, with a view to demonstrating benefits with respect to mortality, stroke and the development of AF.

Apart from stabilizing the atria electrically, atrial pacing may suppress premature atrial contractions that can trigger some episodes of PAF. Dual‐site pacing is another approach, where pacing the right atrium and coronary sinus or nearer the ostium may reduce the frequency of AF but many patients still require concomitant anti‐arrhythmic therapy.82 Another method is interatrial pacing at the posterior triangle of Koch for the prevention of recurrent AF, and one report83 suggests that this technique is safe, feasible and associated with less arrhythmic episodes.

Implantable atrial defibrillator

The implantable atrial defibrillator or atrioverter has been developed for the treatment of frequently recurrent persistent AF, and also for patients with PAF of long duration.84 The treatment is only delivered after AF has begun by delivering a low‐energy synchronized shock. Whilst this therapeutic option appears to be effective, safe and reliable, experience with this device is still limited at the moment.

Radiofrequency ablation

Another method to control the ventricular rate in AF is by using radiofrequency ablation. Techniques include atrioventricular (AV) node ablation with implantation of mode‐switching dual‐chamber pacemaker, ablation of accessory pathways in pre‐excitation syndrome with AF, modulation of AV node, or ablation of AF. AV nodal ablation and implantation of mode‐switching dual‐chamber pacemaker implantation often improve symptom scores and quality of life measurements in the majority of patients.85,,86 However, ‘ablate and pace’ for PAF appears to be associated with a higher progression to persistent AF.86 The most recent meta‐analysis by Wood et al.87 of 21 studies with a total of 1181 patients with refractory atrial tachyarrhythmias (AF in 97%) reported that ‘ablate and pace’ improved quality of life, ejection fraction, exercise duration and healthcare usage, with 1‐year total and sudden death mortality rates of 6.3% and 2.0%, respectively. However, the recent PA3 study found that DDDR pacing compared with VDD pacing did not prevent PAF after atrioventricular junction ablation, with similar times to first AF episode in both groups and an increase in permanent AF within the first year after ablation.88

Experimental studies using linear radiofrequency catheter ablation lesions suggest a high efficacy for these procedures.89,,90 Preliminary human studies have also shown a higher success rate when linear ablations are performed in the left atrium than in the right atrium.91 However, these ‘catheter maze’ procedures are prolonged, it is difficult with present technology to achieve a linear conduction block consistently, and there are major concerns regarding thromboembolic sequelae. Ablation of a specific arrhythmogenic focus responsible for the PAF has been reported to cure some patients with PAF.26,27,91,,92 In such patients, these foci in the pulmonary veins firing repeatedly may initiate paroxysms of AF, and such foci are very amenable to catheter ablation. However, ablation of pulmonary vein foci may carry a risk of pulmonary vein stenosis. In patients with paroxysms of AF preceded by episodes of atrial flutter on Holter monitoring, eradication of the flutter circuit by ablation therapy may also abolish the episodes of AF.93


Surgery for AF is a major procedure, and the therapy of last resort in AF. The two main techniques are the ‘corridor’ and ‘maze’ procedures. The so‐called ‘corridor’ procedure isolates the fibrillating atria from a strip of tissue connecting the sinus and the AV nodes. The limitation of the ‘corridor’ procedure is the persistence of AF in one or part of the atrial chambers, and the haemodynamic and thromboembolism problems remain unresolved. The ‘maze’ procedure 94–,96 attempts to abolish AF by channelling the atrial activation impulse between a series of incisions. Surgery is effective, but is associated with a high incidence of sinus node dysfunction. Both procedures are major, and are associated with a high morbidity, although the current vogue is to combine the procedure with mitral valve surgery.


Paroxysmal AF has been rather neglected in the past, due to its diverse definition and lack of appreciation of its importance despite the associated morbidity and stroke risk comparable to that of chronic AF. Furthermore, the diverse clinical presentations of PAF makes the development of generalized strategies for its management difficult. While there has been a tendency to treat PAF in a similar way to sustained AF, it is now realized that treatment objectives may be different for PAF, with more emphasis on prevention of paroxysms and maintenance of sinus rhythm, but still requiring appropriate consideration of antithrombotic therapy. However, management strategies are limited by evidence from small RCTs, with some inconsistencies over the definition of the arrhythmia and the inclusion of (mainly) symptomatic subjects. Evidence for antithrombotic therapy is also based on extrapolation from epidemiological evidence and subgroup analyses of large RCTs. However, ongoing trials should produce some answers, especially regarding the long‐term prognostic merits of rhythm control compared to rate control.


We acknowledge the support of the City Hospital NHS Trust Research & Development programme for the Haemostasis Thrombosis and Vascular Biology Unit.


  • Address correspondence to Professor G.Y.H. Lip, Haemostasis, Thrombosis and Vascular Biology Unit, University Department of Medicine, City Hospital, Birmingham B18 7QH. e‐mail: G.Y.H.LIPbham.ac.uk


View Abstract