Atrial fibrillation and atrial flutter

• Atrial fibrillation (AF) is a chaotic, rapid and irregular electrical activity of the atria. The atria stop contracting effectively, the ventricles beat irregularly, and blood can clot in the left atrial appendage. It is the commonest sustained arrhythmia worldwide.^[1][2]
• Atrial flutter is a single, regular, looping circuit in the right atrium, usually with 2:1 conduction to the ventricles and a ventricular rate near 150 per minute. It carries a raised thromboembolic risk, so anticoagulation follows the same rules as for AF.^[1][5]
• Management has four jobs, done together: treat the cause and risk factors, prevent stroke, control symptoms with rate or rhythm control, and reassess regularly.^[1] An unstable patient gets synchronised cardioversion first.
• Stroke prevention is decided by a risk score (CHA2DS2-VASc, or CHA2DS2-VA in the 2024 European Society of Cardiology (ESC) guideline), not by whether the AF is paroxysmal or permanent. Direct oral anticoagulants (DOACs) are preferred; warfarin is required with a mechanical valve or moderate-to-severe mitral stenosis.^[1][2]
• Aspirin is not an adequate substitute for anticoagulation.^[1][2][3]

By the end of this page you should be able to:

1. Define AF and atrial flutter, classify AF by pattern, and recognise both on an ECG.
2. List the triggers and risk factors, and explain how AF arises and causes atrial thrombus.
3. Manage acute AF: decide who needs emergency cardioversion, and choose rate or rhythm control with correct drugs and doses.
4. Plan safe cardioversion, including the 24-hour versus 48-hour question and the anticoagulation rules.
5. Calculate CHA2DS2-VASc, weigh bleeding risk, choose and dose an anticoagulant, and name the patients who must take warfarin.

AF is an atrial arrhythmia with uncoordinated atrial activation. On the surface ECG it shows as the absence of discernible, regular P waves and an irregular response of the ventricles, which means there is no pattern to the RR intervals provided the atrioventricular (AV) node is conducting normally.^[1] The atrial rate during AF is about 300 to 500 beats per minute, which the AV node filters down to a much slower, irregular ventricular rate.^[2]

A diagnosis needs an ECG record: a standard 12-lead ECG (10 seconds) or a single-lead device tracing of 30 seconds or more. Smartwatch readings need ECG confirmation.^[1]

AF tends to progress as the atria remodel; the 2023 American Heart Association and American College of Cardiology (AHA/ACC) guideline frames it as staged, from risk factors to permanent AF, so prevention begins before the first episode.^[2]

AF found only by an implanted device (a pacemaker, defibrillator or loop recorder) is called device-detected or subclinical AF. It carries a lower stroke risk than clinical AF, and the benefit of anticoagulation is less certain; the benefit shown in trials applies to ECG-confirmed clinical AF.^[1][2]

• "Chronic AF" and "lone AF" are obsolete; use the pattern names above and look for every risk factor.^[2]
• "Valvular" and "non-valvular" AF are retired. The DOAC trials were called "non-valvular" yet allowed most native valve disease and bioprosthetic valves, excluding only moderate-to-severe mitral stenosis and mechanical valves.^[1][2] Name the lesion instead: a mechanical valve or moderate-to-severe (usually rheumatic) mitral stenosis needs warfarin; other valve disease, such as mitral regurgitation or aortic stenosis, can take a DOAC.^[1]
• "Secondary AF" is now called trigger-induced AF: AF that first appears with a reversible precipitant such as sepsis. Patients still usually have an underlying susceptibility, so they get the full assessment.^[1]

The circuit. Typical flutter is a macro-re-entrant circuit: an electrical wave that chases itself round a large loop of tissue, here around the tricuspid valve in the right atrium. The loop must pass through the cavotricuspid isthmus (CTI), a narrow strip of atrial muscle between the tricuspid annulus and the mouth of the inferior vena cava. That is why it is called CTI-dependent flutter.^[2][5]

• Typical, counter-clockwise flutter is the commonest form. The wave travels up the septum and down the free wall. The ECG shows continuous "sawtooth" flutter (F) waves, negative in the inferior leads (II, III, aVF) and positive in V1.^[2][5]
• Reverse typical (clockwise) flutter uses the same loop in the opposite direction, so the polarity is inverted: positive F waves in the inferior leads and negative in V1.^[2][5]
• Atypical flutter is any flutter that does not use the CTI, such as loops around scar from previous surgery or ablation, or around the mitral valve. The old "type I and type II" names are no longer used.^[2]

The rate. The atrial rate in typical flutter is roughly 240 to 350 beats per minute (cycle length 170 to 250 ms).^[5] The AV node cannot conduct every impulse, and the usual response is a 2:1 block: every second flutter wave reaches the ventricles.^[5] An atrial rate near 300 with 2:1 conduction gives a regular ventricular rate near 150 per minute. This is a useful bedside clue: a regular narrow-complex tachycardia fixed at about 150, which does not slow with exercise or position, should make you look for flutter waves.^[5] Temporarily slowing AV conduction (for example with a vagal manoeuvre) can reveal them.^[2] Variable block gives an irregular rhythm that can be mistaken for AF.^[5]

Class IC antiarrhythmic drugs (flecainide, propafenone) slow the atrial circuit, and without an AV-nodal blocker the AV node may then conduct 1:1, producing a dangerously fast ventricular rate.^[1][5]

Why ablation works so well. The loop depends on one narrow, fixed strip of muscle, so a single ablation line across the CTI that blocks conduction in both directions breaks the circuit. Radiofrequency ablation achieves block in about 92% of patients with one procedure and 97% with repeat procedures (meta-analysis of 21 studies), and CTI ablation beat antiarrhythmic drugs in small randomised trials.^[1][5] Ablation is therefore often offered after the first symptomatic episode, but 50 to 70% of patients develop AF in follow-up, so they need continued review.^[1]

Worldwide, an estimated 59.7 million people were living with AF in 2019, and prevalence in US community cohorts reached up to 5.9%. The lifetime risk for older individuals is up to 1 in 3, and the number affected keeps rising with population ageing, better survival from other heart disease, more risk factors and better detection.^[1] Women with AF are generally older and more symptomatic than men.^[1]

Atrial flutter has an incidence of about 88 per 100,000 person-years, rising to 317 per 100,000 above age 50; more than half of patients with flutter go on to develop AF.^[1]

Holiday heart syndrome is new AF or flutter in an otherwise healthy person after a bout of heavy drinking, named after cases around holidays and weekends. Chronic heavy drinking raises arrhythmia risk too; one review associates it with intake above 36 g of alcohol a day.^[6] The person should still be assessed fully, because the episode may have exposed an underlying tendency.

• Age and hypertension. Incidence rises steeply with age (lifetime risk up to 1 in 3), and hypertension is a leading modifiable risk factor through left ventricular hypertrophy and left atrial enlargement.^[1][2]
• Heart failure. AF and heart failure promote each other (see heart failure).^[1]
• Valve disease, particularly mitral stenosis and regurgitation, coronary disease, cardiomyopathies and congenital heart disease.^[4][1]
• Obesity and obstructive sleep apnoea. Both predict recurrence after cardioversion or ablation, and weight loss of 10% reduced AF burden in a randomised trial.^[1]
• Diabetes (about 25% of patients with AF), chronic kidney disease, chronic lung disease, inflammatory diseases and cancer.^[1]
• Alcohol (dose-dependent), physical inactivity, smoking, and a family history.^[1][2]

1. Triggers. Most AF starts with rapid ectopic firing from muscle sleeves that extend into the pulmonary veins at the back of the left atrium. This discovery, by Haïssaguerre and colleagues, is why ablation isolates the pulmonary veins.^[2] Abnormal calcium handling in atrial cells promotes the afterdepolarisations that fire these triggers.^[2]
2. Substrate and re-entry. Interstitial fibrosis, stretch and inflammation make conduction uneven. A wave that meets patches of slow and blocked conduction can loop back on itself (re-entry). Persistent AF needs a large enough or heterogeneous enough substrate to sustain many such circuits.^[2]
3. Remodelling. AF changes the atria so that AF becomes easier. Electrical remodelling shortens the atrial action potential and reduces gap junctions, structural remodelling brings fibrosis and dilatation, and these changes cause the progression from paroxysmal to persistent AF.^[2] This is sometimes summarised as "AF begets AF".
4. Loss of atrial contraction. In AF the atria quiver without producing a coordinated contraction.^[1] The "atrial kick", which normally tops up ventricular filling late in diastole, is lost, which matters most when the ventricle is stiff or the rate is fast.^[1]
5. Irregular, rapid ventricular response. The AV node transmits a random selection of the atrial impulses. A persistently fast rate can weaken the ventricle over weeks to months, causing tachycardia-induced cardiomyopathy, which can improve once the rate or rhythm is controlled.^[1]
6. Stasis and thrombus. With no effective contraction, blood pools in the left atrium, especially in the left atrial appendage, a blind-ended pouch. Atrial dilatation, endothelial changes and a prothrombotic state add to it. Thrombus formed here can embolise to the brain or other arteries. Thrombus forms mainly in the appendage, and mitral stenosis adds to the risk.^[1][2]

About 90% of patients with AF describe some symptoms, though even in symptomatic patients some episodes go unnoticed.^[1] Whether the patient has symptoms does not predict stroke, embolism or death.^[1]

• Palpitations: a fast, irregular, "flopping" or racing heartbeat.
• Fatigue and reduced exercise tolerance: more common than palpitations, and a major cause of poor quality of life.^[1]
• Breathlessness, orthopnoea or ankle swelling, from the loss of atrial contraction and fast rate. This is especially likely when the ventricle is already stiff or weak.
• Light-headedness, presyncope or syncope, from a fast rate with low output, or from long pauses when AF ends (tachycardia-bradycardia syndrome; see bradyarrhythmias and AV block).
• Chest pain, from rate-related ischaemia in coronary disease.
• Presentation with a complication: stroke, transient ischaemic attack (TIA) or an embolus to a limb or gut, or heart failure, in a person with no previous symptoms.

The modified EHRA (European Heart Rhythm Association) score grades symptom burden and is useful before and after treatment.^[1]

Ask for the cause: alcohol, drugs and caffeine; thyroid symptoms; rheumatic fever, murmurs or valve surgery; hypertension, diabetes and heart failure; snoring; recent infection or surgery; family history.

• Pulse. Irregularly irregular, with varying volume.^[4] Listen for the murmurs of mitral stenosis and regurgitation, and look for raised jugular venous pressure and signs of heart failure.
• Blood pressure and perfusion, to decide whether the patient is stable.
• Look for causes: goitre and eye signs, signs of alcohol excess, and signs of lung disease or sepsis.

Haemodynamic instability due to AF means the arrhythmia is causing shock (low blood pressure with poor perfusion), acute pulmonary oedema, ongoing ischaemic chest pain, or a falling level of consciousness. These patients need immediate synchronised cardioversion (see management).^[1][3]

Multifocal atrial tachycardia is managed along the lines of AF, but beta-blockers are best avoided in chronic lung disease.^[2] Regular narrow-complex tachycardias are in supraventricular tachycardias. The wider approach to a fast rhythm is in approach to tachyarrhythmias.

Investigations ask: is it AF, what caused it, how high are the stroke and bleeding risks, and is it safe to cardiovert? The 2024 ESC basic work-up applies to every patient.^[1]

• 12-lead ECG. Confirms the diagnosis, shows the rate, and may show ischaemia, pre-excitation, or a QT problem before antiarrhythmic drugs. See ECG interpretation.
• Blood tests: full blood count (anaemia, infection), kidney function and electrolytes (drug dosing, potassium, magnesium), liver function, glucose or HbA1c, and thyroid function (TSH).^[1]
• Transthoracic echocardiography: valve disease (especially mitral stenosis), left ventricular function (which decides rate-control drugs), left atrial size, hypertrophy, and right heart function. The National Institute for Health and Care Excellence (NICE) recommends it when a baseline is needed, rhythm control is planned, or structural disease is suspected.^[1][3]

• Ambulatory ECG (Holter, patch or event recorder) for paroxysmal AF, for AF burden and for checking rate control.^[1]
• Exercise ECG to check rate control on exertion, or before and during class IC drugs.^[1]
• Troponin and NT-proBNP when ischaemia or heart failure is suspected, and these can refine stroke and bleeding risk.^[1]
• Transoesophageal echocardiography (TOE) to look for left atrial appendage thrombus before an early cardioversion, or to assess valves when transthoracic images are inadequate.^[1][3]
• Coronary imaging, cardiac MRI and brain imaging when indicated by the story.^[1]

The 2024 ESC guideline organises care into AF-CARE: Comorbidity and risk-factor management, Avoiding stroke and thromboembolism, Reducing symptoms by rate and rhythm control, and Evaluation and dynamic reassessment.^[1] Reassessment is routine because AF and the patient change over time. In plain steps:

1. If unstable, restore sinus rhythm by synchronised cardioversion at once.
2. Find and correct triggers and risk factors.
3. Decide anticoagulation by risk score.
4. Choose rate control, rhythm control, or both.
5. Review the plan regularly.

Synchronised electrical cardioversion is the first-choice treatment when AF causes acute or worsening haemodynamic instability, and anticoagulation should not delay it.^[1][3] The shock is synchronised, which means the machine delivers it just after the R wave, so that it does not fall on the vulnerable part of the T wave and trigger ventricular fibrillation.

Which energy applies? Three figures appear across MedGrasp and each belongs to a different situation: in unstable AF the Resuscitation Council UK gives 120 to 150 J (see Supraventricular tachycardias); in pre-excited AF the European Resuscitation Council advises the maximum output (see Wolff-Parkinson-White syndrome); and for planned elective cardioversion a biphasic shock of at least 200 J is used, as below.

In practice:^[1]

• Sedate with intravenous midazolam, propofol or etomidate (or use anaesthesia), with blood-pressure and oximetry monitoring.
• Use a biphasic defibrillator, which is more effective than monophasic. Pad position matters little: anterior-posterior and antero-lateral positions gave similar results. Pressing on the pads and using the maximum fixed energy both improved success compared with low, escalating energies.^[1] For elective cardioversion of AF, a biphasic initial shock of at least 200 J can improve first-shock success (class 2a).^[2]
• Early relapse is common in sick patients with unfixed triggers, so treat the cause as well.^[1]
• Amiodarone is a slower alternative or adjunct.^[1]

Pre-excited AF (WPW). All the rate-control drugs, and intravenous amiodarone, are contraindicated.^[1] The AHA/ACC 2023 guideline adds: unstable pre-excited AF gets electrical cardioversion; in stable patients, IV ibutilide or IV procainamide is an alternative to elective cardioversion; verapamil, diltiazem, amiodarone, digoxin, adenosine and beta-blockers cause harm; and catheter ablation of the accessory pathway is recommended.^[2] See Wolff-Parkinson-White syndrome.

Most patients do not need to choose between the two. The ESC puts it this way: most need a combination, which should be consciously re-evaluated at follow-up, and rhythm control should be considered in every suitable patient.^[1] This section sets out each approach and then compares them.

Rate control slows the ventricular response by blocking the AV node, leaving the atria fibrillating. It is the standard sole strategy when symptoms are mild and sinus rhythm is not pursued, and it is a bridge in every acute presentation.^[1]

How fast is fast enough? Lenient control, aiming for a resting heart rate under 110 beats per minute, was as good as strict control (under 80 at rest and under 110 on exercise) for clinical events in the RACE II trial of permanent AF. Lenient control is therefore an acceptable first target, unless symptoms persist or a rate-induced cardiomyopathy is suspected.^[1][2]

Choosing the drug. The decision turns on left ventricular ejection fraction (LVEF).^[1]

• LVEF above 40%: beta-blockers, diltiazem, verapamil or digoxin are all acceptable first choices.
• LVEF 40% or lower: beta-blockers and/or digoxin. Diltiazem and verapamil are contraindicated because they depress contractility and can worsen heart failure.
• Acute setting: beta-blockers (any LVEF) or diltiazem/verapamil (LVEF above 40%) act faster than digoxin. Avoid combining a beta-blocker with diltiazem or verapamil except under close monitoring. In haemodynamically unstable patients or those with severely impaired LVEF, IV amiodarone, landiolol or digoxin may be used.
• NICE: in suspected acute decompensated heart failure, seek senior advice about beta-blockers and avoid calcium channel blockers.^[3]

Digoxin acts more slowly and controls exertional rate poorly, so it suits sedentary patients and can be added in heart failure. If one drug fails, NICE advises combining two of a beta-blocker, diltiazem and digoxin.^[3]

When drugs fail. "Ablate and pace" (pacing the ventricle, then ablating the AV node) gives a slow, regular rate regardless of atrial activity.^[1]

Rhythm control tries to restore and hold sinus rhythm, using cardioversion, antiarrhythmic drugs, catheter or surgical ablation. It improves quality of life when sinus rhythm is maintained.^[1] Older trials found no survival benefit over rate control, and drug-based strategies sometimes increased hospital admissions.^[1]

That picture has shifted. In EAST-AFNET 4, rhythm control begun within a year of diagnosis reduced a composite of cardiovascular death, stroke and hospitalisation compared with usual care; in selected patients with reduced ejection fraction, catheter ablation reduced mortality and hospitalisation (CASTLE-AF), while the broader CABANA trial showed no significant difference.^[1]

Features favouring an attempt at rhythm control include symptoms, suspected tachycardia-induced cardiomyopathy, a short AF history, a non-dilated left atrium, and patient preference.^[1]

Cardioversion can be electrical or pharmacological. The risk is that restoring sinus rhythm can dislodge or expose atrial thrombus, and the atrium is temporarily stunned (it contracts poorly) after sinus rhythm returns, so stroke risk is concentrated around the procedure.^[2]

Who needs to wait? The old "48-hour rule" says that AF of definite onset under 48 hours can be cardioverted without prior anticoagulation, while AF of 48 hours or more (or unknown duration) needs preparation. NICE and the AHA/ACC guideline still use 48 hours.^[2][3] The 2024 ESC guideline decided that safety should come first and lowered the line to 24 hours, because the true onset is often not known, and because thromboembolic risk does not appear to be uniformly low in the first 48 hours.^[1][2] In every guideline, if the duration is unknown, treat it as long.

For AF beyond that time, there are two routes:^[1][2][3]

1. Anticoagulate first. At least 3 weeks of uninterrupted therapeutic anticoagulation before the procedure (reliable DOAC adherence, or INR 2.0 or higher on a vitamin K antagonist). Rate control is provided during this time.
2. TOE-guided cardioversion. If the cardioversion cannot wait, a transoesophageal echocardiogram looks for left atrial and appendage thrombus. If none is seen, cardioversion goes ahead under therapeutic anticoagulation started as soon as possible. NICE considers the TOE-guided and conventional approaches equally effective and suggests TOE if a minimal period of anticoagulation is wanted because of patient choice or bleeding risk.
3. If thrombus is found: postpone, anticoagulate for at least 4 weeks (ESC) or 3 to 6 weeks (AHA/ACC), and repeat the TOE before cardioverting.

After cardioversion, continue anticoagulation for at least 4 weeks in everyone, even if the CHA2DS2-VA score is 0; the ESC makes this optional only if the onset was definitely under 24 hours and risk is low. Long-term anticoagulation is then decided by the risk score, not by whether sinus rhythm is restored.^[1][2]

Electrical or pharmacological? Electrical cardioversion is more effective, but needs sedation or anaesthesia. NICE recommends electrical cardioversion when AF has lasted longer than 48 hours.^[1][3] Pharmacological cardioversion needs no sedation and suits stable patients with recent-onset AF, but spontaneous conversion occurs in 76 to 83% (69% within 48 hours), and in RACE 7 ACWAS waiting up to 48 hours was non-inferior to immediate cardioversion.^[1]

NICE offers flecainide or amiodarone to people with no structural or ischaemic heart disease, and amiodarone to those with structural heart disease.^[3]

"Pill in the pocket". A selected patient with infrequent, brief, symptomatic paroxysmal AF and no structural heart disease may carry a single oral dose of flecainide or propafenone to take at onset. Safe use requires screening for sinus node disease, AV conduction defects and Brugada syndrome, and a test dose given first in hospital.^[1]

Antiarrhythmic drugs reduce AF recurrence by 20 to 50% and do not eliminate it; a recurrence is not treatment failure if episodes become less frequent, shorter or milder.^[1] Choose the drug for safety first, and tailor it to the heart:^[1]

• Heart failure with reduced ejection fraction: amiodarone (monitor for extracardiac toxicity).
• Heart failure with mildly reduced or preserved ejection fraction (HFmrEF, HFpEF), ischaemic heart disease, valvular disease: dronedarone (400 mg twice daily).^[2] It must not be given with NYHA class III-IV heart failure or an episode of decompensated heart failure in the previous 4 weeks, and must not be used for rate control in permanent AF.^[1][2]
• Structurally normal heart, no coronary disease or severe hypertrophy: flecainide (typically 50 to 150 mg twice daily) or propafenone (150 to 300 mg three times daily), with a beta-blocker, diltiazem or verapamil to prevent 1:1 flutter.
• Sotalol (80 to 160 mg twice daily if creatinine clearance is over 60 mL/min, once daily at 40 to 60, contraindicated under 40) may be considered with normal LVEF or coronary disease; it needs close QT, potassium and renal monitoring and has a mortality signal.^[2]
• Avoid antiarrhythmic drugs in advanced conduction disease unless a pacemaker is in place. Do not combine antiarrhythmics.

Amiodarone causes thyroid dysfunction in 15 to 20% of patients, and may cause lung, liver, skin and nerve toxicity, so it is a last choice in younger patients and is avoided for long-term rate control.^[1][2][3]

Catheter ablation uses radiofrequency heat or cryoenergy (freezing), or newer pulsed field ablation (high-voltage pulses), delivered through a catheter in the left atrium to isolate the pulmonary veins, the source of the triggers. Pulmonary vein isolation is the cornerstone of AF ablation.^[1]

• Indication. Recommended for symptomatic paroxysmal or persistent AF that does not respond to, or cannot tolerate, antiarrhythmic drugs. It is also a recommended first-line option after shared decision-making in paroxysmal AF, with similar adverse event rates to drugs. It is less clear that first-line ablation is better than drugs in persistent AF.^[1] NICE: consider ablation if drugs fail or are unsuitable.^[3]
• Heart failure with reduced ejection fraction. In selected patients, ablation can improve ejection fraction and outcomes.^[1]
• Results and risks. Serious complications are uncommon but real, and operator experience matters. It is not a guaranteed cure, and recurrences are managed with drugs or a repeat procedure.^[1][3]
• Surgical ablation (the Maze family) is considered at the same time as other heart surgery in symptomatic patients.^[1][3]

Patients remain on anticoagulation after ablation according to the risk score, not the result of the procedure.^[1]

AF raises the relative risk of ischaemic stroke to about 2.3.^[1] Anticoagulation removes much of that risk: vitamin K antagonists and direct oral anticoagulants reduce stroke substantially compared with placebo or no treatment.^[1] This section covers who to treat, the bleeding side of the balance, which drug to use, and what to do when things go wrong.

The score gives one or two points to each of the following risk markers. Age is counted once, in the band that applies (so the maximum score is 9).^[1][2]

How the guidelines use it. All agree on the principle that the score, not the AF pattern, decides.^[1][2] They differ in the details:

• ESC 2024 (CHA2DS2-VA): anticoagulation is recommended with a score of 2 or more and should be considered with a score of 1, through shared decision-making. The sex criterion was dropped because female sex acts as a modifier that depends on age, rather than a risk factor in itself, and because it complicates practice. Anticoagulation is recommended regardless of score in AF with hypertrophic cardiomyopathy or cardiac amyloidosis.^[1]
• AHA/ACC 2023: uses estimated annual thromboembolic risk. Anticoagulation is recommended at 2% a year or more (for example, CHA2DS2-VASc of 2 or more in men or 3 or more in women), and is reasonable at 1% to under 2% (CHA2DS2-VASc 1 in men, 2 in women).^[2]
• NICE 2021: uses CHA2DS2-VASc. Offer a DOAC at a score of 2 or more; consider one for men at 1. Do not offer anticoagulation to people under 65 whose only point is female sex (a score of 0 in men or 1 in women).^[3]

The score is modest in its predictive power; other risk modifiers include cancer, chronic kidney disease, biomarkers (troponin, BNP) and atrial enlargement. Borderline cases call for shared decisions.^[1][2]

Assess bleeding risk to find and correct modifiable factors, not to deny anticoagulation. The modifiable factors are uncontrolled hypertension, labile INR on warfarin, concomitant antiplatelet or NSAID use (and SSRIs, per NICE), excess alcohol, and reversible causes of anaemia.^[1][3]

The HAS-BLED score is the best known bleeding tool. It counts uncontrolled hypertension (systolic over 160 mmHg), abnormal renal or liver function, stroke, bleeding history or predisposition, labile INR, elderly age (over 65 in the original score) and drugs or alcohol.^[2] The guidelines' attitude to it has changed:

• ESC 2024 recommends against using bleeding risk scores to start or withdraw anticoagulation, because their predictive ability is modest and under-treatment causes strokes.^[1] AHA/ACC notes that HAS-BLED discriminates poorly and shares risk factors with stroke.^[2]
• NICE recommends the ORBIT score, used to guide discussion and risk reduction, not as a cut-off.^[3]

Absolute contraindications are few (for example a primary intracranial tumour, or intracerebral bleeding related to cerebral amyloid angiopathy), and most others are temporary or relative.^[1] Do not withhold anticoagulation solely because of age or a risk of falls.^[3]

The four DOACs (apixaban, dabigatran, edoxaban, rivaroxaban) are preferred to vitamin K antagonists in eligible patients. They lower stroke and intracranial haemorrhage at least as well as warfarin and are simpler to take, with fewer interactions and no INR monitoring.^[1][2]

Warfarin (a vitamin K antagonist, VKA) is required for:

• Mechanical heart valves. In a trial of dabigatran against a vitamin K antagonist in mechanical valves, there were more thromboembolic and major bleeding events with dabigatran and the trial stopped early; a trial of apixaban after mechanical aortic valve implantation was stopped for excess thromboembolic events.^[1]
• Moderate-to-severe mitral stenosis, usually rheumatic. In 4531 patients with rheumatic heart disease and AF, warfarin-type drugs gave fewer cardiovascular events and deaths than rivaroxaban, without more bleeding.^[1]

Bioprosthetic valves (including mitral) and valves after transcatheter aortic valve implantation are not restricted: a DOAC is acceptable.^[1] Warfarin also remains an option for patients who cannot afford DOACs or who take interacting drugs such as enzyme inducers.^[2]

Practical points:

• Use the full tested dose. Under-dosing a DOAC is common and increases stroke without reducing bleeding. Reduce only when the labelled criteria are met.^[1][2]
• Stop most antiplatelets when starting a DOAC unless there is a clear separate indication.^[1]
• Warfarin control. The usual INR target is 2.0 to 3.0 (mechanical mitral valves and some aortic mechanical valves need a higher target set by the valve guideline^[8]), and a TTR above 70% is the benchmark. Switch to a DOAC if TTR is under 70% and the patient is eligible; NICE advises reassessment if TTR is under 65%.^[1][3] In frail patients over 75 with good control, switching from a VKA to a DOAC led to more bleeding in one trial (17.8 against 10.5 events per 100 patient-years), so a stable patient may stay on warfarin.^[1]
• Reassess regularly. Stroke and bleeding risks change with age and illness.^[1][3]

Minor bleeding often needs only a brief interruption while the source is treated. Major bleeding needs supportive care, stopping the anticoagulant, finding the source (for example by endoscopy), and reversal for life-threatening or critical-site bleeding.^[1]

• Warfarin: prothrombin complex concentrate (PCC) works faster and with fewer complications than fresh frozen plasma, and fresh frozen plasma acts faster than vitamin K. Many protocols add intravenous vitamin K to sustain the effect once the INR has been corrected.^[1]
• Dabigatran: idarucizumab fully reverses the anticoagulant effect; dialysis also removes dabigatran. Activated charcoal or gastric lavage may help if the last dose was within 2 to 4 hours.^[1]
• Apixaban, edoxaban, rivaroxaban: andexanet alfa reverses factor Xa inhibitors. In ANNEXA-I (530 people with a bleed into the brain on a factor Xa inhibitor) it controlled haematoma growth more often than usual care (67.0% against 53.1%), but thrombotic events were more frequent (10.3% against 5.6%).^[9][10] The US Food and Drug Administration later judged the risks to outweigh the benefits, and sales in the US ended on 22 December 2025.^[11] Where an antidote is unavailable, four-factor PCC is often used, with evidence limited to observational studies.^[1]
• Restart is possible once bleeding has stopped and the source has been investigated and treated.^[1]

Thrombus in AF tends to form in the left atrial appendage,^[1][2] so closing it with a plug-like device (percutaneous LAAO, for example Watchman or Amulet) or by surgery at the time of cardiac surgery aims to prevent embolism without long-term anticoagulation.^[1] Its place is limited: the 2024 ESC guideline says LAAO may be considered when there is a contraindication to long-term anticoagulation, and the AHA/ACC guideline calls it reasonable in patients at moderate to high stroke risk with a contraindication due to a non-reversible cause.^[1][2] NICE restricts LAAO to people who cannot have anticoagulation because of a contraindication or intolerance, and does not offer it as an equal alternative.^[3] Device-related thrombus occurs in 1.7 to 7.2%, and antithrombotic treatment is still needed afterwards.^[1]

• After an ischaemic stroke: trials found no benefit from heparin or anticoagulants in the first 48 hours, so low-dose aspirin is given in that window. For starting a DOAC, the ELAN trial (2013 patients) and the TIMING trial (888 patients) found early starting (within about 48 hours after a minor or moderate stroke, or day 6 to 7 after a major one, in ELAN; day 4 in TIMING) no worse than later starting. A specialist neurology team co-ordinates the decision.^[1]
• After bleeding into the brain: there is not enough evidence to say whether to start or restart anticoagulation. In the small SoSTART trial, starting an anticoagulant did not prove non-inferior to avoiding it (recurrent bleeding 8% against 4%), and APACHE-AF showed no difference in stroke or vascular death with apixaban. The advice is an individual decision led by neurology.^[1]

AF in pregnancy raises the risk of death, so care is shared between obstetric, anaesthetic, neonatal and cardiology teams. Stroke risk is scored as in non-pregnant women.^[1]

• Anticoagulation: unfractionated or low molecular weight heparin is preferred because it does not cross the placenta. Vitamin K antagonists are avoided in the first trimester and from week 36, and DOACs are not recommended.^[1]
• Rate control: a beta-1 selective blocker is first choice, but not atenolol, which can slow fetal growth. Digoxin is the next choice. Verapamil is avoided in the first trimester.^[1]
• Rhythm control: electrical cardioversion is recommended for haemodynamic instability or pre-excited AF, and is safe for the fetus, whose heart rate is monitored. Intravenous ibutilide or flecainide may be tried in a stable woman with a structurally normal heart.^[1]

• Stroke and systemic embolism. Relative risk of ischaemic stroke about 2.3.^[1]
• Heart failure. AF raises the relative risk of heart failure four- to five-fold.^[1]
• Cognitive decline, dementia and ischaemic heart disease (adjusted hazard ratio 1.39 and relative risk 1.61).^[1]
• Death. Up to a twofold increase in all-cause and cardiovascular mortality compared with sinus rhythm.^[1]
• Tachycardia-induced cardiomyopathy, usually reversible with rate or rhythm control.^[1]

The outlook depends largely on underlying disease, whether the stroke-prevention and risk-factor measures are carried out, and how well the rate and symptoms are controlled. Even on anticoagulation, patients keep a residual risk of death, which underlines the need to treat comorbidity.^[1] After post-operative AF, the risk of recurrent AF over the next 5 years is four to five times higher, and the condition is a marker for later stroke, heart failure and death.^[1]

AF-CARE builds in reassessment: review symptoms, the heart rate, stroke and bleeding risk, kidney function (for DOAC dosing), drug interactions, and weight, blood pressure, alcohol and sleep apnoea. Watch for progression to persistent AF, new heart failure, and the effect of treatment on quality of life.^[1] Re-check anticoagulation when age reaches 65, and when diabetes, heart failure, vascular disease or stroke appear.^[3] Patients after flutter ablation should still be reviewed for AF.^[1]

Practical points where resources are limited:

• Diagnosis needs only a pulse check and a 12-lead ECG.
• Warfarin with INR is a safe, cheap and effective strategy when monitored (INR 2.0 to 3.0 unless a mechanical valve sets a higher target, ideally with a TTR above 70%).^[1][8] If INR monitoring is unavailable, make sure the patient and family know the warning signs and the drug interactions (traditional medicines, antibiotics and alcohol).
• Rheumatic mitral stenosis and AF should be treated with warfarin, as warfarin-type drugs did better than rivaroxaban in the INVICTUS trial.^[1]
• Rate control with a beta-blocker, or digoxin if the patient is in heart failure. Check potassium and renal function before digoxin. Avoid diltiazem and verapamil if heart failure is possible and no echocardiogram is available.^[1]
• Cardioversion without TOE or reliable anticoagulation is reserved for the unstable patient; otherwise follow the 3-week rule.^[1]

Whether someone with atrial fibrillation or flutter may drive depends on whether the rhythm has caused, or is likely to cause, sudden incapacity, and on the national licensing authority. Advise the patient not to drive after any episode of collapse or near-collapse until the rhythm is assessed and treated, and tell them to check the rules that apply to them. The UK DVLA rules for arrhythmia are set out in Supraventricular tachycardias (Driving section) and apply in the same way here.

• Haemodynamic instability (while resuscitating and cardioverting).
• AF with heart failure, a murmur, or a suspected valve problem not already assessed.
• Symptoms despite adequate rate control, or a young patient, for rhythm control or ablation.
• Atrial flutter, for consideration of ablation.
• AF with pre-excitation.
• Doubt about the safest anticoagulant (mechanical valve, pregnancy, very high bleeding risk, kidney failure, recurrent stroke despite anticoagulation).
• Syncope with AF, or long pauses when AF ends.