Atrial fibrillation

Atrial fibrillation is a part of "Cardiac arrhythmias",  a Level 1 topic from Section 2.1.3 in the second edition of the CICM Syllabus for the Second Part Examination. It is very common in the ICU, reflecting in the increasing number of past paper SAQs which discuss the fibrillating atrium, and what precisely you ought to do about it.

• Question 18 from the first paper of 2024
• Question 9 from the second paper of 2021
• Question 14 from the first paper of 2012
• Question 2 from the second paper of 2004.

The arrhythmia chapter of Oh's Manual (Management  of  cardiac  arrhythmias  by Andrew  Holt in the now largely historical version that gathers dust on the shelf of the author) was the original and most instinctive reference for this topic. The chapter is a marvellous resource, consisting of highly concentrated information compacted in layers upon layers into a dense sedimentary deposit. Nuggets of gold are embedded within.  Most of the end section (pages 216 to  224) deals almost exclusively with atrial fibrillation, and this summary chapter uses it and its references ad libitum to answer the SAQs. As far as useful  information goes, the 2023 AHA guidelines and the 2024 ESC guidelines can be viewed as the definitive documents for management and investiations, in spite of the fact that its intended audience is largely ward-based and outpatient physicians. As is pointed out below, AF in the ICU is a completely different animal to "outpatient AF".  

Epidemiology of atrial fibrillation in the ICU

It is mostly likely a complete accident that the CICM exam question asking for six risk factors came approximately three months after the publication of AFIB-ICU,  a multicentre cohort study describing that exact answer for that exact cohort.  The incidence of AF was 15.6% (13.3% was completely new). One might be tempted to expand on this by including a list of patient categories most likely to develop AF, but that would sound a lot like a list of risk factors, and the list of risk factors runs very close to becoming a list of causes, for which there is a whole dedicated section below.

Taking a slightly broader view, in Australia AF was listed as a diagnosis in 1.9% of total hospital presentations that one year, and has a community prevalence of something like 1.5-5% in the population of the Antipodes, which is apparently among the highest in the world. The lifetime risk is something like 25%, i.e. one in four of us will get AF at one stage or another.  We sure do love to fibrillate, especially when we are in the ICU; and especially when we are in my ICU, where my excellent colleague Sumesh Arora did his college project on AF in the dismal dark ages of 2007, finding an incidence of 29.5% (and we didn't even have cardiac surgery at that stage). Considering the incidence of AF in single-centre audits varies from low single digit percent (4.1%, in Californian trauma patients,  to 46% among the septic Taiwanese), this number, as well as all other numbers, is entirely meaningless. The reader should finish this paragraph with the vague impression that AF is common in ICU. 

Atrial flutter vs. atrial fibrillation

Flutter deserves a brief aside as a detour along the road to meatier topics.  Yes, both flutter and AF are atrial arrhythmias, but totally different pathology is involved in each. Flutter is an intra-atrial re-entrant circuit which just goes around and around (usually, counter-clockwise and in the right atrium). It is apparently notoriously drug-resistant, and in ye olde Oh's Manual, Andrew Holt recommends ibutilide, a Class III antiarrhythmic, as the drug most likely to cardiovert this rhythm, which is baffling because this drug is not approved for use in Australia. But then judging by the tone of Holt's chapter, there is nothing to recommend any specific agent anyway. A group is listed in a nonchalant manner ("Drugs such  as  digoxin,  diltiazem,  beta-adrenergic blockers, sotalol  and  amiodarone  may  be  tried;  the  choice depends on LV function"). We are warned against the use of Class 1A and 1C agents, as they may result in 1:1 conduction (and a ventricular rate in excess of 200). By comparison, the more recent article by Herzog et al (2017) is much more firmly prescriptive. It separates the management of flutter into four main areas, described by the RACE acronym: Rate control, Anticoagulation, Cardioversion and Electrophysiology. The article summarises the recommendations from the most recent guidelines from the AHA/ACC:

• Same drugs for rate control as AF (but higher doses)
• Same anticoagulants as AF
• Cardioversion: chemical choices include ibutilide or amiodarone
• Cavo-tricuspid isthmus ablation (EPS)

Generally, rate control is more difficult to achieve than with AF. UpToDate authors suggest calcium channel blockers (diltiazem or verapamil) but the level of evidence is somewhat sub-standard ("we prefer", the authors say). The problem with ICU patients is that frequently, the (higher) dose required to slow AV nodal conduction would produce an intolerable level of toxicity (i.e. the LV would not be grateful for the decreased contractility). Ergo, amiodarone becomes the natural choice in the ICU, because it covers both the "R" and the "C" of the RACE algorithm. 

Aetiology of atrial fibrillation in ICU

Question 14 from the first paper of 2012 asked the candidates to list several non-cardiac causes of AF. The list of causes could be long, and it is possible to organise it in several ways. One way is the familiar VINDICATE framework which should theoretically make it easier to memorise the causes. A strategy which might be more useful (actually enhancing one's practice) would be to organise the causes according to their pathophysiology. Both are made available below.

Causes of Atrial Fibrillation Organised by System

Vascular: Myocardial infarction Pulmonary embolism Pulmonary hypertension Subarachnoid haemorrhage Infectious: Sepsis Myocarditis Pericarditis Infective endocarditis Neoplastic: Cardiac mass, eg. myxoma Drug-induced: Catecholamines Alcohol Caffeine Idiopathic: Infiltrative disease, eg. amyloidosis Age-related fibrotic changes

Idiopathic: Infiltrative disease, eg. amyloidosis Age-related fibrotic changes Congenitial: Atrial septal defect Familial AF Autoimmune: Autoimmune myocarditis Traumatic: Cardiac contusion Cardiac surgery Endocrine/environmental: Hypothermia Hyperthyroidism Haemochromatosis/iron overload Phaeochromocytoma Electrolyte derangement

Causes of Atrial Fibrillation Organised by Pathophysiology

Catecholamine excess Exogenous (eg. adrenaline infusion) Endogenous: Subarachnoid haemorrhage Stress Phaeochromocytoma Thyrotoxicosis (indirectly) Atrial distension Pulmonary hypertension: Obstructive sleep apnoea Pulmonary embolism Primary pulmonary hypertension Pulmonic valve disease Septal defects Valvular disease, including infective endocarditis

Abnormality of conducting system Congenital cardiac disease, eg. septal defect Infiltrative cardiac disease, eg. amyloidosis Ischaemic heart disease Age-related fibrotic changes Haemochromatosis/iron overload Hypothermia Increased atrial automaticity / irritation Alcohol Caffeine Catecholamines Electrolyte derangement Myocarditis

Of the risk factors, the most dependable is age. All data seems to point to an increased risk of AF, from around 0.4% in the under-60s to something like 12% in the over-75s, according to Cavaliere et al (2006). 

Pathophysiology and clinical course of AF in ICU

So those are the associated causes, but how do they cause AF?  Two main factors are required:

1. An "arrhythmogenic atrial substrate"
2. An initiating event

That "substrate" is basically anything that causes the atria to become damaged, distended, or vaguely irritated - hich is a long list (basically all of the contents of the tables above). The basic features are:

• Atrial structural remodelling (eg. fibrosis, anything that damaged the atria chronically like valvular disease)
• Atrial electrical remodelling (eg. tachycardia, chronic inflammation, etc - anything that causes changes in ion channel expression)

The initiating event is then something else that comes along and triggers an atrial cell focus to develop automaticity. These include:

• Increased sympathetic, or decreased parasympathetic drive (eg. hypovolemia, or the use of some kind of sympathomimetic)
• Electrolyte abnormalities (hypokalemia, hypomagnesemia)
• Cardiac ischaemia
• Bacterial endotoxin during sepsis
• Acute mechanical stress (eg. volume overload, or just being poked with a guidewire)

Assessment of AF in ICU

"Outline your assessment" is a tired workhorse of the exam process, dragging the plough through these following familiar furrows:

• History
• Examination
• Investigations (which can be further subdivided into Imaging and Biochemistry)

The google search strategy "assessment of AF" yields too many primary care articles. For the unique aspects of investigating AF in ICU, specific recommendations are in order, and Bosch et al (2018) have some excellent points:

History for the assessment of AF in ICU

The objectives of taking a "history" (i.e. interrogating the bedside nurse and inspecting the progress notes) would surely have to be mostly to do with finding risk factors and initiating factors. The upshot of this is that the next few points will seem like a pointless duplication of already mentioned material. Their presence here is mostly to facilitate an effortless cut-and-paste into some future "outline" SAQ answer, should the opportunity present itself.

• History of major risk factors for AF in the medical background (see the massive tables above)
• History of increased sympathetic, or decreased parasympathetic drive :
  • Shock, vasoactive agent support
  • Pain, agitation
  • Increased effort or exercise, eg. tachypnoea
• History suggestive of electrolyte abnormalities (eg. recent use of diuretics, fasting, surgery, fluid resuscitation, vomiting or diarrhoea, etc)
• History that might hint at cardiac ischaemia (eg. chest pain)
• History of sepsis
• Recent acute mechanical stress to the atria (eg. recent line insertion, pleural effusion, volume resuscitation)
• Administration of arrhythmogenic drugs 

Additionally, one may wish to branch out into risk stratification:

• Duration of observed monitored AF, and history of palpitations (suggestive of chronicity, which affects your decisions to cardiovert or anticoagulate)
• CHADSVASC2 factors in history (see below)
• Risks for bleeding, considering the possibility of future anticoagulation
• History of syncope, suggestive of occasions where the AF was haemodynamically significant

Examination findings for assessment of AF in ICU

Examination in this situation ends up focusing on the causes and consequences of the AF, rather than the AF itself, which is often a rather unimpressive physical finding (the pulse is irregular). One really needs to get inventive here. For example:

• Neurological examination to reassure onself that there is no stroke
• Findings classical for thyrotoxicosis or toxidromes
• Cardiovascular examination to look for features of heart failure, shock, valve abnormalities, and any suggestion that there might be vascular embolic phenomena
• Auscultation to seek effusion, infection, oedema
• Pleural rub suggestive of PE
• Abdominal examination for pain of gut ischaemia

These are long stretches, and for the majority of patients, will be mostly negative. All of the money is in imaging and biochemistry:

Imaging and biochemistry for the assessment of AF in ICU

The ICU patient is often unable to narrate the history, the progress notes are often incomplete or cut-and-pasted, and the bedside staff are just relieving while the usual nurse is having a meal break. Fortunately, the ICU patient is often well investigated, and there will be objective data to assess:

• Biochemistry:
  • Electrolytes, especially potassium and magnesium
  • TFTs 
  • acid-base findings on the ABG results 
  • Recent cardiac and inflammatory biomarkers
• ECG: neither imaging nor biochemistry, but still an investigation? Mostly looking for ischaemia and characteristic findings of specific electrolyte abnormalities
• Imaging:
  • CXR,  to look for chronicity (eg.  bronchial splaying suggestive of a large LA), causes (eg. pleural effusion) or consequences (eg. pulmonary oedema)
  • TTE, looking for atrial remodelling
  • TOE, obviously, looking for atrial appendage clot

How much do the electrolytes matter? In a word, a lot.  Rafaqat et al (2022) looked this up and down, in case the readers want a detailed discussion of exactly how they matter. The bottom line is that electrolyte derangement (and especially low electrolyte values) seem to be associated with an increased incidence of AF in all populations. 

Management of atrial fibrillation in a broad sense

"Outline your initial management of the tachycardia", asks Question 14 from the first paper of 2012. Treatment for AF is a pursuit of several goals: control of ventricular rate, protection from systemic emboli and restoration of sinus rhythm. The best destination for the reader who wishes to appear for the CICM Second Part Exam sounding as if they know their job would have to be the 2023 AHA guidelines and the 2024 ESC guidelines

Rate control vs. rhythm control

Holt's chapter for Oh's Manual mentions four studies (AFFIRM, RACE, STAF and PIAF) in support of rate control instead of rhythm control. They were published between 2000 and 2003.  A more recent meta-analysis (Caldeira et al, 2012) identified four more (a total of eight) suitably high-quality studies, featuring data from 7499 patients. The mortality data from these was underwhelming. "No clear survival benefit is apparent", laments Holt. Rate control was found to be superior only in terms of the composite endpoint (death, stroke and recurrent hospitalisation). Theoretically, rhythm control should actually be better (particularly in the absence of significant structural heart disease) because it may prevent myocardial remodelling due to AF.

Methods of rate control

Multiple methods are available. Oh's Manual lists essentially the entire antiarrhythmic arsenal,  cautioning against the use of digoxin in patients with enhanced sympathetic tone (useless, apparently). Single-dose flecainide is apparently also good, but in patients with structural heart disease it tends to cause sudden cardiac death.

The 2014 AHA statement gives the following recommendations:

• β-blocker or calcium channel blocker for paroxysmal AF, which can be given IV if they are haemodynamically stable (CCBs are favoured in COPD patients)
• IV amiodarone for critically ill patients
• If the patient has some sort of pre-excitation problem, these AV node blockers are not recommended (cardioversion is the better choice; the AHA also recommend procainamide for stable patients but this is not available in Australia)
• If the patient is in decompensated heart failure, calcium channel blockers are not recommended (use digoxin or amiodarone instead). In the long term these people seem to benefit from a combination of digoxin and a β-blocker

They recommend a rate of 80 or so as the endpoint to aim for, but give a slightly weaker recommendation in favour of a more "lenient" rate (~110) provided the LV function is well-preserved. 

Methods of rhythm control

For the haemodynamically unstable patient, DC cardioversion is the gold standard. The arrhythmia management algorithm in the pre-arrest management section of the ARC ALS Handbook  (based on ILCOR Guidelines) recommends this approach. For the haemodynamically stable patient, there are a variety of chemical cardioversion options, among which amiodarone and vernakalant (the latter being a novel drug with rapid activity against AF and  a reasonably benign side-effect profile). Electrical cardioversion may still be an option, within 48 hours of onset or later following adequate anticoagulation and TOE.

Systemic anticoagulation

The options are:

• Warfarin: relative risk reduction for stroke 62%; absolute risk reduction 2.8% per year
• Aspirin: relative risk reduction for stroke 22%; absolute risk reduction 1.5% per year
• Warfarin plus aspirin: no additional benefit over warfarin alone
• Dabigatran: 35% reduction in stroke compared to warfarin

The 2014 AHA statement recommends:

• No anticoagulation is a reasonable choice if the AF is within 48 hours and the CHA₂DS₂-VASc score is 0; the risk of stroke is 0.2%.
• If the AF is within 48 hours and the score is anything but 0, IV heparin is recommended.
• If the AF has been going on for longer than 48 hours (or, god knows how long) - IV heparin should be used.
• If you are going to anticoagulate, anticoagulation with something should continue for at least 3 weeks before and 4 weeks after their TOE-cardioversion.
• Aspirin is an alternative if the score is 1
• Warfarin or similar if the score is 2 or greater
  • Essentially this means that if you're over 75 you automatically score some warfarin because that age bracket immediately gets you a score of 2 from the CHA₂DS₂-VASc scoring system.

Risk stratification

The CHA₂DS₂-VASc scoring system is the recommended method of determining the risk of stroke. In essence it comes down to three main categories: score 0, score 1 and any score of 2 or more.

C	Congestive heart failure (or Left ventricular systolic dysfunction)	1
H	Hypertension: blood pressure consistently above 140/90 mmHg (or treated hypertension on medication)	1
A2	Age ≥75 years	2
D	Diabetes Mellitus	1
S2	Prior Stroke or TIA or thromboembolism	2
V	Vascular disease (e.g. peripheral artery disease, myocardial infarction, aortic plaque)	1
A	Age 65–74 years	1
Sc	Sex category (i.e. female sex)	1

A score of 0 is hard to get, but confers a virtually negligible risk of stroke (~ 0%). A score of 1 equates to a risk of 1.3% and a score of 2 puts you in a high risk category (2.2%). The maximum score is 9, with an associated stroke risk of 15.2%. Mind you, these are annual risks. What's the daily risk in ICU patients? Nobody knows. Sibley et al (2015) mulls the problem over and concludes that whatever that risk is, it surely mist be higher in the ICU population, particularly among patients with sepsis. At present there does not appear to be either any sort of scoring system or any sort of guidelines statement to help direct the intensivist here, apart from generic guidelines like the 2014 AHA statement.

Management of AF in the ICU

There are several important differences between "free-range" AF in the movie-going public and "ICU AF " in the critically ill patient.

• AF in ICU is a transient phenomenon. Only about 15% of the fibrillating ICU patients will still be fibrillating at the time of their discharge from the ICU (Artucio et al, 1990).
• The cause of AF in ICU is usually non-structural, reversible, and non-cardiac (i.e. related to the cause of the non-cardiac critical illness). A large epidemiological survey of non-cardiothoracic ICU patients by Kanji et al (2012) confirmed that most of the time AF in the ICU can be attributed to such causes as shock, sepsis, electrolyte disturbance, atrial stretch, hypoxia, inotropes, and so forth.  All of these are problems which can be expected to resolve at the end of the critical illness. 
• The mortality and morbidity in the ICU population are different to those of the general fibrillating population. The ICU patient with AF is less likely to die of stroke, but is more likely to develop worsening multi-organ system failure due to poorly controlled shock (when they lose that atrial systolic kick). Among ICU patients it is difficult to say whether the AF is independently associated with increased mortality, or whether it is merely a marker of greater illness severity.
• Given that "ICU AF" is likely to be associated with shock and a catecholamine excess, it makes sense that a beta-blocker may be the most appropriate choice of drug. However, choice of agents is clouded by methodologic heterogeneity among trials. Kanjii et al (2008) made valiant attempts to perform a meta-analysis of known methods, and came up against a wall of wildly differing methodologies and outcome measures. Standards are lacking. Unstable patients were excluded from all studies (which makes them completely useless in the ICU patients who are unstable almost by definition).
• As far as chemical cardiversion goes, Kanjii et al (2008) reported on some figures. Esmolol and flecainide seem to give the best chance of cardioversion (60-100% success rate at 12 hours), with amiodarone far behind (40-76% success rate at 12 hours). Even magnesium was better than amiodarone (51-93% success rate).

Therefore, the management strategy should focus on controlling the cause, and preventing the complications of a rapid rate, with rhythm control only considered for specific populations who cannot afford to fibrillate (eg. severe mitral regurgitation) or those in whom the cause has been and gone (eg. post operative patients who have recovered from their arrhythmogenic surgery). 

How to structure this? A good effort by Johnston et al (2022) guided what follows.

Addressing the treatable causes of AF

This obviously depends on the causes. It is also essential, as the AF will remain for as long as the cause persists. It is also rather difficult to summarise this in a written exam answer and still sound like a human, as opposed to an AI-generated response. One may consider something like "directed strategies to control the modifiable risk factors", but this is so vague and nebulous that it could not possibly score any marks. "Treat sepsis, cardiac ischaemia and respiratory failure" is more specific but likely to be too specific as not all AF in ICU is due to these (merely most of the AF). It is probably safest to just say "treat the treatable causes" without getting too academic. 

Minimising arrhythmogenicity

There are a host of factors imposed by the ICU environment which are not exactly causes, per se, but are contributors to the AF. Some can  be modified to decrease the overall tendency of atria to fibrillate, which can be broadly described as "stop irritating them". Reducing the arrhythmogenic agents, eg. weaning the inotropes or reducing the diuretics, is likely to be the highest yield strategy. These could be correcting electrolyte deficits, removing mechanical stimuli (eg. the PA catheter or the pleural effusion) and reducing the sympathetic tone by managing pain and agitation, reversing the shock, replacing volume, correcting hypercapnia or acidosis, and extubating (or, reintubating) the patient.

Pharmacological rate control

β-blockers are first line, as these are recommended for inpatient rate control by all the societies, and are the least toxic. Calcium channel blockers are the next alternative. For those patients in ICU who cannot tolerate the nerfed inotropy, the next in line are amiodarone vernakalant and ibutilide, of which amiodarone is definitely the most familiar and best studied. For patients with a normal LVEF, flecainide is also an option. Digoxin is listed by AHA as an option for "sedentary" patients, but in the ICU, the sympathetic drive is often too strong to make this a viable solo agent. Digoxin is usually seen co-administered with something else to make it more effective. 

Endpoints for managing AF in the ICU

Rate, rather than rhythm, is the goal of management, unless the patient falls into one of the categories listed below under "selection for cardioversion". The various royal societies recommend slightly different things: AHA wants a heart rate of  <110 for most, and <80 for those with reduced LVEF, whereas the most recent ESC guidelines favour flexible individualised goals. The numbers are probably meaningless and the intensivist often finds themselves mostly concerned about the pragmatic need to avoid tripping the mandatory observation alerts in the ward, and will therefore aim for whatever rate is safely below those. Where patient care is actually the main focus, most people would agree that the endpoint rate is whatever rate below which the patient experiences no palpitations, has no cardiac ischaemia, and is not in overt heart failure. 

Selection for cardioversion

All major guidelines agree that haemodynamically unstable patients will benefit from cardioversion, but in ICU, haemodynamic stability is relative (we can make the blood pressure whatever we want), risks of sedation may be great (by definition, the patient is after all in ICU) and the reversion to sinus rhythm is short lived (until the cause is corrected). Thus, only some patients might be suitable, and this is reflected in the findings of AFIB-ICU, who had about 4% of their cohort undergo cardioversion. What makes those lucky few so special? Niederdoeckl et al (2022) looked at pharmacological cardioversion and Shima et al (2021) looked at electricity to see what makes a successful cardioversion candidate. In short, the following features made for a higher likelihood of success:

• Young patients
• Concomitant use of an antiarrhythmic
• Short duration of AF
• Flutter, as opposed to AF

Additionally, the following features favour cardioversion on grounds other than the likelihood of success:

• The stimulus is gone: eg. the patient is recovering from cardiac surgery
• Normal TTE: most likely to stay sinus (whereas large dilated atria are unlikely to return to normal function)
• Already sedated and ventilated: safe, from the patient's perspective
• Severe acute mitral regurgitation or diastolic failure: these are patients who cannot affort to be in AF for haemodynamic reasons
• Extremely high bleeding risk (eg. patients recently experiencing a intracerebral haemorrhage) - i.e. those who cannot afford to be anticoagulated

Which brings us to:

Anticoagulation for the non-cardioverted

Anticoagulation is the natural strategy for long term protection of patients who cannot get back into sinus rhythm, and is the mainstay of management in the community, where it reduces the risk of stroke over years. The pragmatic intensivist may not care overmuch as to how great the benefit might be, but for interest's sake, let's observe the numbers:

• For patients with a maximum CHA₂DS₂-VASc score (9 points), the stroke risk is 12.2% per year (or, 17.4% risk of stroke/TIA/systemic embolism). Overall, AF patients as a cohort seem to have a risk of stroke about 2.5 times higher than the general population, who have a risk of stroke something like 0.18% per year.
• According to Menichelli et al (2021), the overall risk of ischaemic stroke in anticoagulated AF patients (i.e. those with a score of 2-3 points) is about 0.99%, whereas without anticoagulation it is 2.5%. On the basis of this, assuming some sort of mathematical continuity, the higher risk patients will also have their stroke risk reduced proportionally, i.e. from 12.2% down to something like 4.9%. But there is no data to confirm this.
• An ICU stay is rarely measured in years, and so the intensivist must contend with a daily risk of stroke, which would therefore be something like 0.0027% for a low-risk patient, or 0.033% for the highest risk category, which is about 0.23% per week.
• The upshot of this is that anticoagulation is often not indicated in the acute setting, or too risky to implement for the unclear benefits. The risk of bleeding and the difficulty titrating even the short acting agents is considerable. Nelson et al (2021) estimated that something like 17-48% of reported AF patients in ICU receive anticoagulation, and then about 7-8% report major bleeding complications, whereas the incidence of thromboembolic events seems to be at most about 1.4%, irrespective of whether you are anticoagulated or not. "The little evidence available does not support therapeutic anticoagulation for NOAF whilst patients are critically ill", concluded Miller et al (2022), looking at basically the same data.

Consequences of atrial fibrillation in ICU

Why do we even try to control AF?

• Adverse haemodynamic effects due to the following factors:
  • loss of atrial systole (the "kick")
  • Decreased diastolic filling time during rapid AF
  • "Tachycardiomyopathy" - a global cardiomyopathy associated with the rapid rate
• Systemic embolism and stroke

What could go wrong? AFIB-ICU found that patients who developed AF had more risk of anticoagulation-related bleeding (5.9% vs 2.7%), but also thromboembolic events (13.6% vs 7.9%), which means some either did not get anticoagulation or that it was ineffective. Most papers also seem to report an effect on increasing mortality, though in all honesty it would be very difficult to adjust the statistics accurately in a way that accounts for the presence of the original critical illness (it was sepsis wot killed im, not the fibrillation)