Rate control: Ventricular rate control remains a major therapeutic target for AF in HF patients. Beta-blockers and digoxin (as adjunctive therapy) are the main agents available for systolic HF. In addition, non-dihydropyridine calcium channel antagonists (verapamil, diltiazem) can be used instead of beta-blockers in HF with preserved EF. Finally, amiodarone can be considered for rate control if combination of beta-blocker and digoxin is inadequate[41]. A number of studies have demonstrated prognostic benefit of beta-blockers in AF with HF. A retrospective analysis of the US Carvedilol HF trial data focused categorically on patients who had AF at the time of enrolment. In comparison to the placebo arm, the beta-blocker group had improved LV ejection fractions and better physician-determined global assessment. Moreover, there was a tendency towards reduced combined cardiovascular mortality and hospitalisation[42]. The Digitalis Investigation Group trial showed that although digoxin did not affect mortality in HF, it reduced the number of hospital admissions. AF was among the exclusion criteria and as such these results may not be applicable to AF in HF[43]. Moreover, digoxin loses its effect during periods of catecholamine excess and is not recommended as monotherapy. Of note, a recent post-hoc analysis of the the Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) trial has cast doubt on the safety of digoxin in HF patients[44]. It was shown that digoxin is associated with increased all-cause mortality including a 41% increased risk of death in patients with CHF or LVEF of less than 40%. This should, however, be interpreted with caution as AFFIRM was designed to compare rate and rhythm control and patients were not randomised to digoxin therapy. Moreover, only 25% of the AFFIRM cohort had HF. Moreover, a propensity matched analysis of the same cohort failed to demonstrate any increase in mortality with digoxin. It is likely that the patients on digoxin in the study had higher risk of mortality[45]. Finally, data on the use of verapamil and diltiazem in HF is limited[46]. Verapamil has been shown to be useful in HF patients with normal LV systolic function[47] and current ESC guidelines recommend their use in HF with preserved ejection fraction as an alternative to beta-blockers[41]. However, these should be avoided in HF with reduced ejection fraction due to negative inotropic effect on LV contractility[41,48].

Another point that needs further clarification relates to the optimal target ventricular rate for permanent AF patients in HF. Rate control efficacy in permanent Atrial fibrillation: a Comparison between lenient vs strict ratE control II (RACE II) trial looked at lenient (110 bpm) vs strict (< 80 bpm) ventricular rate control in patients with AF. The primary end-point was a composite of cardiovascular death, HF admission, bleeding and embolic events including stroke. No significant difference was observed in the two arms[49]. Again less than 35% of the cohort had HF (15% in NYHA class 4) and results may not be generalizable to the HF patients. Routine versus Aggressive upstream rhythm Control for prevention of Early atrial fibrillation in HF (RACE III) is currently recruiting and will provide definitive answers for the HF population[50].

Rhythm control: Amiodarone and dofetilide are the main anti-arrhythmic agents assessed in HF patients with AF. survival trial of antiarrhythmic therapy in congestive heart failure (CHF-STAT) and atrial fibrillation and congestive heart failure (AF-CHF) trials have demonstrated the efficacy of amiodarone in cardioversion and maintenance of sinus rhythm in patients with moderate to severe LV systolic dysfunction[51,52]. Its overall effect on mortality was shown to be neutral but long-term clinical use remains limited due to a risk of significant side effects. DIAMOND-CHF trial looked at the effect of dofetilide on a cohort of mainly ischaemic HF patients. A pooled sub-study analysis incorporated 506 patients who were in AF. Dofetilide was shown to be safe with an overall neutral effect on mortality. It was superior to placebo in cardioversion and patients on the drug were more likely to be in sinus rhythm at one year as compared to placebo (79% vs 42%). Moreover, it was also associated with reduced HF admissions[53]. Importantly, patients who converted to sinus rhythm had lower all-cause mortality (in the dofetelide as well as placebo arms) signifying the beneficial prognostic impact of sinus rhythm. However, torsade de pointes (1.6%) remains a cause for concern with dofetilide and requires initiation in hospital under close monitoring. Furthermore, it is not available in Europe. Subsequently, dronedarone (an iodine-free amiodarone derivative) was introduced with a promising adverse effect profile. A post hoc analysis of the ATHENA (A Trial with dronedarone to prevent Hospitalization or death in patiENts with Atrial fibrillation) looked at stable patients with LVEF less than 40% and NYHA 2-3 symptoms. It showed a reduced risk of all-cause mortality and/or hospitalisation due to cardiovascular events[54]. However, ANtiarrhythmic trial with dronedarone in Moderate-to-severe congestive heart failure Evaluating morbidity DecreAse (ANDROMEDA) trial (which looked at patients with severe HF in sinus rhythm and not AF) had to be terminated prematurely because dronedarone increased mortality in such patients[55]. As a result, it is no longer licensed for use in patients with unstable/severe HF.

Rate vs rhythm control: There is no convincing scientific evidence so far to support a rhythm control strategy in preference to rate control. Given the negative impact of AF in HF, the concept of maintaining sinus rhythm appears attractive, yet a number of randomised trials have failed to demonstrate improved long-term outcomes with a rhythm control approach[56-59]. The results are, however, limited by the fact that these trials were not exclusive to HF (for instance only 25% of the AFFIRM cohort had depressed LV function) and it may be difficult to apply these findings to the HF population. On the other hand, there have been trials looking exclusively at HF patients as well. AF-CHF enrolled 1376 patients with systolic HF. They were randomized to either rhythm or rate control and followed up for 3 years looking at prospective data for mortality, HF admissions and stroke[52]. The difference in cardiovascular mortality observed in the two arms was not significant (27% in rhythm control vs 25% in rate control respectively). It is noteworthy, however, that only 80% of patients in the rhythm control arm remained entirely AF-free (65% when looking at overall 3 year follow up visits as well as the 21% who crossed over to rate control arm)[11]. Interestingly, a post hoc analysis of the AFFIRM trial looked at the rhythm control arm of the trial. Sinus rhythm was associated with less severe NYHA symptomatic class and better functional capacity (assessed by 6-min walk test)[60]. Similarly, in a subgroup analysis of CHF-STAT trial, Kaplan-Meier analysis of the survival curves for those who converted to SR with amiodarone showed significantly better survival as compared to those who remained in AF[51]. Same conclusion can be derived from the DIAMOND sub-study as well[53]. However, these results are based on post-hoc subgroup analyses and should be applied with caution. A recent meta-analysis of the 4 main randomised control trials of AF rate vs. rhythm control in HF (incorporating 2486 patients) has demonstrated no significant difference in terms of mortality and thromboembolic events[61].

Thromboprophylaxis: Although beyond the scope of this review, it would be amiss not to mention the enormous clinical, social and economic impact of stroke in HF patients with AF. Due to various co-morbidities, patients with HF have a significantly higher risk of thromboembolic events particularly stroke. Hence, oral anticoagulation is imperative unless there are equally binding contraindications. ACCF/AHA/HRS guidelines have kept the option of either aspirin or anticoagulation for patients with a CHADS₂ score of 1 while European Society of Cardiology (ESC) and Carbon capture and storage (CCS) guidelines indicate anticoagulation for such patients in preference to aspirin. Nevertheless, there is unanimous agreement in recommending long-term anticoagulation for all patients with a CHADS₂ score of 2 and above[62]. Warfarin is well recognized in this regard and has been the mainstay of thromboprophylaxis in AF[63] for the last 60 years. It has been shown to reduce the risk of stroke by as much as 65% and is thrice as efficacious as aspirin[64]. Its clinical utility, however, is fraught with a variety of limitations (both real and perceived, by patients and physicians alike). These include a narrow therapeutic window, need for meticulous monitoring of INR levels, unreliable blood levels due to interaction with various drugs/food and risk of bleeding in an increasingly frail/ ageing population.

The last few years have witnessed the exciting development of a novel group of oral anticoagulants (NOACs) with the advantage of rapid onset of action, fewer drug/food interactions and predictable blood levels thus exonerating patients from laborious INR monitoring. They have been shown to carry a lesser risk of intracranial bleeding in comparison to warfarin while maintaining the same level of protection against stroke. However, widespread use is restricted by higher costs, unavailability of a reversal agent in the event of a major bleed and no validated lab markers of anticoagulant effect[65]. The two main classes consist of direct thrombin inhibitors (dabigatran) and activated factor X inhibitors (apixaban, rivaroxaban, edoxaban) while several others are under development. Dabigatran was the first to be approved by Food and Drug Administration in 2010 for non-valvular AF following the Randomized Evaluation of Long-term anticoagulation therapY (RE-LY) trial[66] which enrolled 18113 patients with AF. One-third of the study population had symptomatic HF or LVEF < 40%. Patients were randomized to receive either 150 or 110 mg twice daily (blinded dose groups) of dabigatran or INR-guided warfarin therapy. In comparison to warfarin, 110 mg twice daily dose was non-inferior in efficacy and superior in safety while the 150 mg twice daily dose was superior in efficacy and had similar rates of major bleeding. Consequently, dabigatran has been recommended as an alternative to warfarin in recent ESC, AHA/ACCF as well as CCS guidelines[67-69]. Similarly, Rivaroxaban was studied in the Rivaroxaban Once daily oral direct factor Xa inhibition Compared with vitamin K antagonism for prevention of stroke and Embolism Trial in Atrial Fibrillation (ROCKET-AF) looking at over 14000 patients. Data demonstrate non-inferiority to warfarin in terms of efficacy. It was associated with less intracranial haemorrhage as well albeit a higher risk of gastrointestinal bleed[70]. Finally, Apixaban is the only one so far which has been shown to be superior to warfarin in reducing the primary end-point of thromboembolic events including stroke (annual event rate 1.27% vs 1.60%; P < 0.001 for noninferiority; P = 0.01 for superiority). This is derived from the Apixaban in Preventing Stroke and Systemic Embolism in Subjects With Nonvalvular Atrial Fibrillation trial which enrolled over 18000 patients. The 21% reduction in primary safety end-point was mainly derived from a lower likelihood of haemorrhagic strokes. There was no significant difference in the rates of ischaemic strokes between the two[71]. All three NOACs available so far have been licensed for use in non-valvular AF.

Drug therapy is the mainstay of AF management. However, many patients are unable to achieve rhythm or rate control targets due to therapeutic inefficacy or side effects respectively. Consequently, device therapy and electrophysiological catheter interventions have gained importance.

“Pace and ablate’’ strategy: Atrio-ventricular node (AVN) ablation accompanied by a permanent pacemaker is often used as an extreme option for definitive rate control. However, AF is not eliminated per se and rate control with a regular RR length may not suffice in compensating for the haemodynamic detriment caused by A-V dys-synchrony and loss of atrial systole. Thus, arguably, the procedure may only be of symptomatic benefit[72]. Moreover, there is a potential for progressive inter-ventricular dys-synchrony due to chronic RV pacing. Hence, cardiac resynchronisation therapy (CRT) has emerged as the pacing option of choice in all patients with systolic HF[73,74] who require pacing for AVN ablation. On the other hand, it is well recognized that clinical response to CRT is hampered if adequate AF rate control cannot be achieved. This is likely to be due to a lower percentage of biventricular pacing and here AVN ablation can be very helpful. This has been demonstrated in a recent meta-analysis of 23 observational studies involving 7495 CRT patients (25% of the total had AF). When compared to patients in sinus rhythm, presence of AF conferred a higher likelihood of CRT non-response and increased all-cause mortality (10.8% vs 7.1% per year, pooled RR = 1.50, 95%CI: 1.08-2.09, P = 0.015). In addition, there was a lesser improvement in quality of life, exercise capacity and LV end-systolic dimensions. On the other hand, in patients with AF, AVN ablation not only improved response to CRT (RR = 0.40, 95%CI: 0.28-0.58, P < 0.001) but was associated with a reduced risk of mortality as well[75]. A number of other small, mostly single-centre, non-randomized studies of CRT (in HF patients with AF) have also shown improvement in soft end-points such as reduced mitral regurgitation, improved LV ejection fraction and better exercise capacity but clearly more data is required[76-78]. For instance, a registry-based analysis of patients with severe HF compared 139 patients who had AF with 445 in sinus rhythm. One year follow up revealed comparable CRT-related improvement in NYHA symptom class and LV ejection fractions in the two cohorts. Of note, mortality was higher in the AF group (12% vs 7%; OR = 1.80, 95%CI: 0.95-3.4)[78]. Although the results are encouraging, yet large scale placebo controlled randomised trials are still required to confirm long term prognostic benefit.

AF ablation: As noted above, “pace and ablate” strategy is effective in controlling the ventricular rate but it does not eliminate AF as such. Also, like all invasive procedures CRT is not free of potential complications. Consequently, radio-frequency catheter ablation (RFA) using pulmonary vein isolation (PVI) has gained momentum in the management of AF. A number of observational studies (albeit small) provide supportive data for such a strategy. The non-randomized observational study by Hsu et al[79] compared 58 patients in HF with an equivalent number of age/sex matched controls without HF. All underwent RFA for AF. At the completion of one year, 78% of the HF cohort and 84% of controls remained in sinus rhythm (although 50% had required a second procedure due to recurrence of AF). RFA led to significantly improved LV function (mean increase 21%) in the HF cohort. In addition, significant improvement was seen in NYHA symptom class, quality of life (assessed by SF-36 QoL scores) and exercise capacity (assessed by bicycle-ergometer stress test) as well. The trial was, however, not powered to look at mortality trends. Similar results have been obtained in a number of other small non-randomized studies demonstrating improvement in LVEF and patient symptoms[80-82]. Pulmonary vein Antrum isolation vs atrioventricular node ablation with Biventricular pacing for treatment of atrial fibrillation in patients with congestive heart failure (PABA-CHF) was a multi-centre study which prospectively randomized 81 drug-refractory AF patients (with a LVEF of 40% or less and NYHA functional class 2-3) to undergo PVI or AVN ablation with biventricular ICD implant. They were followed up at 6 mo. The composite primary end point consisted of LVEF, 6-min walk distance and Minnesota Living with Heart Failure score. PVI patients fared better in all three components of the end point than the cohort who underwent AVN ablation and biventricular pacing[83]. Recently, MacDonald et al[84] conducted a randomised controlled trial in HF patients comparing rhythm control by RFA (n = 22) to rate control by medical therapy (n = 19). RFA failed to show any significant improvement in radionuclide LV ejection fractions as compared to the rate control arm. Only 50% were able to retain sinus rhythm at the end of one year and a significant (15%) complication rate was observed. A meta-analysis of AF ablation trials in patients with moderate LV systolic dysfunction looked at 9 studies involving a total of 354 patients. RFA led to an overall improvement in LV systolic function. However, the results are limited by heterogeneous study cohorts and lack of long-term outcome data[85]. Hence, large scale, multicentre, randomized controlled trials with longer follow up will be required for further definitive clarification. Finally, in patients undergoing cardiac surgery, surgical ablation techniques (variations of Cox Maze procedure) are available as a safe and effective alternative[86] including for those with depressed LV function[87].

Selective AV nodal vagal stimulation (AVN-VS) has emerged as a potentially viable therapeutic intervention for ventricular rate control in AF. Loss of vagal tone followed by sympathetic overstimulation is thought to contribute to the pathophysiology of HF. Epicardial AV nodal fat pad stimulation (using catheter electrodes) targets parasympathetic efferents in the vagal ganglia and confers negative chronotropic and dromotropic effects. This can then be potentially used to modulate AF rate control in patients with HF. Small-scale, randomised preclinical case-control studies have shown effective heart rate control along with improvement in LV function in acute[88] as well as chronic settings[89]. Investigators induced HF and AF in canine models using rapid ventricular pacing for 4 wk followed by continued rapid atrial pacing respectively. Similar reversible negative chronotropic effects have been demonstrated in a cohort of 25 patients who underwent efferent vagal nerve stimulation with a multipolar catheter in superior vena cava or coronary sinus[90]. Although it is only hypothesis generating at this stage, yet it showed consistent slowing of the heart rate and this was associated with improved LV function. Larger trials are needed to ascertain the true potential of this technique.

Currently available anti-arrhythmic agents used for AF act on multiple ion-channels located in the atria as well as ventricles. Consequently, there is a risk of ventricular pro-arrhythmia and this is a particular concern in patients with structural heart disease/HF. Development of atrial specific anti-arrhythmics with a reduced risk of ventricular pro-arrhythmia is indeed an attractive strategy[91]. Vernakalant is a potassium-channel blocker which has undergone successful phase II and III trials. It is different to the conventional class III agents in that it selectively delays atrial repolarization by blocking atrial specific potassium-channels. As a result it suppresses AF by prolonging the atrial refractory period and is not associated with ventricular pro-arrhythmic effects such as QT prolongation and torsades[92]. A phase III superiority study of Vernakalant vs Amiodarone in Subjects With Recent Onset Atrial Fibrillation (AVRO) demonstrated superior efficacy of vernakalant as compared to amiodarone[93]. However, only 20% of the patients in the cohort had HF. Also, there is no experience yet in advanced HF as patients with unstable congestive HF, NYHA class 4 symptoms, or HF requiring inotropes were excluded from the study.

A significant minority of patients in AF are unable to benefit from oral anticoagulation-either due to contraindications (bleeding, allergy) or therapeutic failure (ischaemic stroke despite effective anticoagulation). Studies have shown that in non-rheumatic AF, left atrial appendage serves as the source of thromboemboli in around 90% of cases[94]. Consequently, percutaneous devices for the occlusion/exclusion of the left atrial appendage have emerged as a potentially promising answer to this challenging conundrum[95]. Results from the recent randomised WATCHMAN left atrial appendage system for embolic PROTECTion in patients with Atrial Fibrillation (PROTECT AF) trial have established the feasibility of this technique[96] while demonstrating non-inferiority with warfarin therapy. The initially high rate of procedural complications has subsequently improved with greater operator experience[97] and combination with PVI has been successfully carried out[98] as well. Long term outcome data is not available yet. Trials are also underway assessing further devices such as the Amplatzer cardiac plug and LARIAT suture delivery system[99].

Despite the frequent coexistence of AF and HF, it is intriguing that more than half of even severe HF patients do not develop AF. It is postulated that there may be a genetic predilection for AF in certain HF patients. If such is the case, then modulating these factors may provide a potential therapeutic target. Indeed, familial clustering of AF is well recognized. Moreover, genome wide association studies have demonstrated several common AF-related mutations and polymorphisms[100]. Recently, a large population study showed a strong genetic association between AF and a polymorphism in the ZFHX3 gene (which encodes a cardiac transcription factor). This was associated with increased AF risk in HF patients when compared to the general population[101]. The mechanism by which this translates into pathology is not known. Polymorphisms have also been identified in the beta1-adrenergic receptor gene in patients with systolic HF and AF[102]. Again the exact significance is not clear yet but it may help risk stratify HF patients in terms of favourable response to beta blocker therapy[103].

Apart from ion-channel blockers, other pharmacologic agents have been investigated for potential anti-AF effects with the hope that modification of the arrhythmogenic atrial substrate and neuroendocrine axis may be of benefit. Limited data is available for polyunsaturated fatty acids[104], statin therapy[105] and renin-angiotensin-aldosterone system blockade[106,107]. At best, the findings have been inconclusive so far and larger randomized controlled trials are required[108].