Article Text

Download PDFPDF

Original article
Radiofrequency ablation for persistent atrial fibrillation in patients with advanced heart failure and severe left ventricular systolic dysfunction: a randomised controlled trial
  1. Michael R MacDonald1,
  2. Derek T Connelly1,2,
  3. Nathaniel M Hawkins3,
  4. Tracey Steedman4,
  5. John Payne1,
  6. Morag Shaw4,
  7. Martin Denvir5,
  8. Sai Bhagra1,
  9. Sandy Small2,
  10. William Martin2,
  11. John J V McMurray6,
  12. Mark C Petrie1
  1. 1Golden Jubilee National Hospital, Glasgow, Scotland, UK
  2. 2Glasgow Royal Infirmary, Glasgow, Scotland, UK
  3. 3Aintree Cardiac Centre, Liverpool, UK
  4. 4Western Infirmary of Glasgow, Glasgow, Scotland, UK
  5. 5The Royal Infirmary of Edinburgh, Edinburgh, Scotland, UK
  6. 6BHF Glasgow Cardiovascular Research Centre, Faculty of Medicine, University of Glasgow, Glasgow, Scotland, UK
  1. Correspondence to Dr Michael R MacDonald, Department of Cardiology, Golden Jubilee National Hospital, Beardmore Street, Clydebank, Scotland G81 4HX, UK; michaelrossmacdonald{at}


Objective To determine whether or not radiofrequency ablation (RFA) for persistent atrial fibrillation in patients with advanced heart failure leads to improvements in cardiac function.

Setting Patients were recruited from heart failure outpatient clinics in Scotland.

Design and intervention Patients with advanced heart failure and severe left ventricular dysfunction were randomised to RFA (rhythm control) or continued medical treatment (rate control). Patients were followed up for a minimum of 6 months.

Main outcome measure Change in left ventricular ejection fraction (LVEF) measured by cardiovascular MRI.

Results 22 patients were randomised to RFA and 19 to medical treatment. In the RFA group, 50% of patients were in sinus rhythm at the end of the study (compared with none in the medical treatment group). The increase in cardiovascular magnetic resonance (CMR) LVEF in the RFA group was 4.5±11.1% compared with 2.8±6.7% in the medical treatment group (p=0.6). The RFA group had a greater increase in radionuclide LVEF (a prespecified secondary end point) than patients in the medical treatment group (+8.2±12.0% vs +1.4±5.9%; p=0.032). RFA did not improve N-terminal pro-B-type natriuretic peptide, 6 min walk distance or quality of life. The rate of serious complications related to RFA was 15%.

Conclusions RFA resulted in long-term restoration of sinus rhythm in only 50% of patients. RFA did not improve CMR LVEF compared with a strategy of rate control. RFA did improve radionuclide LVEF but did not improve other secondary outcomes and was associated with a significant rate of serious complications.

Clinical trials registration number NCT00292162.

  • Atrial fibrillation
  • heart failure
  • radiofrequency ablation (RFA)
View Full Text

Statistics from


Patients with heart failure commonly develop atrial fibrillation (AF).1 Loss of atrial contraction, together with the irregular and often rapid ventricular response, leads to a reduction in stroke volume and cardiac output.2–5 The presence of AF in heart failure is associated with increased morbidity and mortality.6–8 Despite this, a strategy of rhythm control has not been shown to be better than one of rate control for improving outcomes in these patients.9 There has been concern, however, that the potential benefit of restoring sinus rhythm may have been offset by adverse effects of antiarrhythmic drug treatment in the clinical trials to date.10 Consequently, there is interest in non-pharmacological approaches to rhythm control, particularly radiofrequency ablation (RFA).

Several non-randomised, observational, studies have reported that sinus rhythm can be restored and maintained by RFA in patients with left ventricular systolic dysfunction, with subsequent improvement in ejection fraction.11–14 However, a strategy of rhythm control using RFA has not been compared with one of rate control in a randomised trial. The aim of our study was to compare the effect of these two strategies on left ventricular ejection fraction (LVEF) in patients with advanced systolic heart failure, severe left ventricular dysfunction and persistent AF.

Patients and methods

The study was approved by the local research ethics committee and registered at Clinical (registration number: NCT00292162). All patients provided written informed consent.


Men and women aged 18–80 years, with New York Heart Association functional class II–IV symptoms despite optimal heart failure treatment for at least 3 months, ejection fraction <35% measured by radionuclide ventriculography, persistent AF and no contraindication to cardiovascular MRI were eligible. Patients were recruited from six heart failure clinics in Scotland between January 2007 and July 2009. The principal exclusion criteria were paroxysmal AF; QRS duration >150 ms (or QRS 120–150 with evidence of mechanical cardiac dysynchrony15); any contraindication to oral anticoagulant drugs; primary valvular disease or acute myocarditis as the cause of heart failure; coronary revascularisation within the preceding 6 months; pregnancy and expected cardiac transplantation within 6 months.

Randomisation and masking

Patients were randomised on a 1:1 basis by computer generated sequence to receive either RFA and medical treatment or medical treatment alone. The treatment allocation was concealed from investigators in numbered envelopes until patients had been through baseline assessments. All scans were analysed by two blinded operators.

Medical treatment

All patients had been receiving optimal heart failure treatment for 3 months. If mean heart rate was >80 bpm over a 24 h period then digoxin was added to treatment.


All patients enrolled were receiving long-term warfarin treatment. If the international normalised ratio was <2 at any time during the month before the procedure, transoesophageal echocardiography was performed before the procedure to exclude thrombus in the left atrial appendage. Warfarin was stopped 2 days before admission to allow the international normalised ratio to fall below 2. Patients were reloaded with 10 mg of warfarin and treated with full-dose low molecular weight heparin 4 h after the procedure. The following morning, before discharge, patients received a second dose of low molecular weight heparin.

Radiofrequency ablation

RFA was performed a median of 43 days from randomisation using the Bordeaux technique.13 All procedures were performed in a single centre, by an experienced operator. A decapolar mapping catheter was advanced into the coronary sinus. After trans-septal puncture, intravenous unfractionated heparin was given to achieve an activated clotting time of 300 s. Pulmonary vein and left atrial anatomy was delineated with pulmonary venous angiography and three-dimensional reconstruction of the left atrium using Nav-X mapping system (St Jude Medical, Minnesota, USA). Pulmonary vein recordings were performed with a duodecapolar variable-diameter circular mapping catheter (Lasso (Biosense Webster, California, USA) or Inquiry Optima (St Jude Medical)). RFA was performed with an irrigated tip ablation catheter (ThermoCool, Biosense Webster). If AF persisted after pulmonary vein isolation, radiofrequency lesions were delivered in a linear fashion between the right and left superior pulmonary veins, and then at sites of complex fractionated atrial electrograms on the interatrial septum, mitral annular region, left atrial roof, left atrial free wall and around the base of the left atrial appendage. In most cases radiofrequency energy was also delivered inside the coronary sinus at sites of complex electrograms. If the patient remained in AF following ablation, sinus rhythm was restored by internal cardioversion under intravenous sedation. If the patient had a history of atrial flutter (or if atrial flutter was seen during the procedure, cavotricuspid isthmus ablation was also performed, and bidirectional isthmus block was confirmed after ablation.

For repeat procedures, RF energy was delivered at sites of pulmonary vein reconnection in order to re-isolate the pulmonary veins; further RF energy was delivered at sites of complex fractionated atrial electrograms as before.

Study follow-up

Study assessments (see below) were made at baseline and after 6 months in those randomised to medical treatment only. In those randomised to RFA, oral amiodarone was started before discharge and continued for 3 months, after which 24 h ambulatory electrocardiographic (ECG) monitoring was conducted. Patients in sinus rhythm returned 3 months later for their final visit assessments (6 months from baseline). Those in AF were scheduled for a second RFA procedure. The median time to this second procedure from the 3 month visit was 90 days. Three months after the second procedure they returned for their final visit.

Study assessments

The following assessments were made at baseline (before randomisation) and at the final study visit: physical examination; a 12-lead ECG; 24 h ambulatory ECG monitoring; measurement of N-terminal pro-B-type natriuretic peptide (NT-proBNP); a 6 min walk test; radionuclide ventriculography and cardiovascular magnetic resonance (CMR). Given the difficulties in measuring LVEF in patients with AF, we opted to measure cardiac function using two techniques: radionuclide ventriculography and CMR.16 It is well recognised that absolute values for ejection fractions measured by CMR and radionuclide ventriculography are markedly different. Health status and quality of life were assessed using the Short Form-36 (SF-36), Kansas City Cardiomyopathy Questionnaire (KCCQ) and Minnesota Living with Heart Failure Questionnaires (MLHFQ).

One additional assessment was made to investigate the time course of any changes in LVEF. All patients returned for a repeat CMR scan early after randomisation. In the RFA group this was 1 week after the initial ablation procedure; in the medical treatment group this was 2 weeks after randomisation.

Cardiovascular magnetic resonance

CMR imaging was performed using a static 1.5 Tesla Siemens Sonata scanner applying protocols previously described.17 In brief, the left ventricular short axis was identified by first piloting a vertical long-axis plane from the transaxial plane. The horizontal long-axis plane was imaged, and from this a stack of short-axis images was obtained during breath-holding, covering the length of the left ventricle. ECG-gated cine images were acquired using a steady-state free precession sequence. Left ventricular end-diastolic volumes (LVEDVs), left ventricular end-systolic volumes (LVESVs) and ejection fraction were assessed offline from the serial short-axis cine loops using dedicated commercially available software (ARGUS, Siemens, Erlangen, Germany). Similarly, left atrial end-diastolic area (LAEDA) was measured using four-chamber cine loops.

Radionuclide ventriculography

Planar radionuclide ventriculography images were acquired using a best septal separation (∼40 left anterior oblique (LAO)) projection and a 70 LAO projection. The images were acquired for 20 min in list-mode format using a GE (General Electric, Connecticut, USA) optima gamma camera with a high-sensitivity parallel hole collimator. List-mode files were subsequently reformatted into a 24-frame gated image using variable-time formatting to include all beats. Analysis was performed after completion of the study so that images could be randomised and analysed by observers blinded to patient details.

Statistical analysis

The prespecified primary end point of the study was change in LVEF measured using CMR from randomisation to the last study visit. With 20 patients per group, the study had 80% power to detect a mean difference in change in ejection fraction of 6.8% between groups (allowing for a 10% drop-out rate and a predicted 80% success rate after two ablation procedures). Secondary end points were change in left and right ventricular ejection fractions (as assessed by radionuclide ventriculography), LVESV, LVEDV, LAEDA, NT-proBNP, 6 min walk and quality of life (assessed using KCCQ, MLHFQ and SF-36). Continuous variables are expressed as mean±SD. Categorical variables are expressed as frequencies or percentage. Ejection fraction and secondary outcomes were assessed as change from baseline to the final visit. Means were compared using an independent t-test (with Levene's test for equality of variances). The primary analysis was performed on an intention-to-treat basis. Post hoc exploratory analyses based on procedural success were also conducted. All statistical tests were two sided and a p value of <0.05 was considered significant. All statistical calculations were performed using Statistical Package for Social Scientists (SPSS), version 16.


Of 366 patients screened, 325 were excluded for reasons shown in figure 1. Of the remaining 41, 22 were randomised to RFA and 19 to medical treatment. Three patients withdrew after randomisation, two of whom were randomised to RFA. Of those randomised to RFA, one withdrew after a stroke that occurred 6 days after his (or her) initial ablation and the second developed an indication for a permanent pacemaker before undergoing ablation. One patient withdrew from the medical treatment arm after the baseline evaluation. Twenty patients assigned to RFA and 18 assigned to medical treatment alone completed follow-up.

Figure 1

CONSORT diagram: enrolment and follow-up of study patients. AF, atrial fibrillation; LVEF, left ventricular ejection fraction; NYHA, New York Heart Association; RFA, radiofrequency ablation.

Baseline characteristics

Approximately 90% of the patients were in New York Heart Association functional class III and almost 80% had been admitted to hospital with heart failure in the past (table 1). The predominant aetiology of heart failure in both groups was coronary heart disease. Both groups had high rates of β-blocker and ACE-inhibitor use and no patients were treated with Vaughan-Williams class I or III antiarrhythmic drugs.

Table 1

Baseline characteristics

Although the groups were well matched for age and sex, there were some imbalances. Those randomised to medical treatment had a longer duration of AF (64 vs 44 months), a higher mean radionuclide LVEF (19.6% vs 16.1%) and smaller LVESV and LVEDV.

Ventricular rate with medical treatment

The average ventricular rate on 24 h ambulatory ECG monitoring at baseline was 75±15 bpm in the RFA group and 72±11 bpm in the medical treatment group. At the final visit it was 69±12 and 73±14 in the RFA and medical treatment groups, respectively.

Radiofrequency ablation

In total, 27 procedures were performed, including six repeat procedures. Two patients required atrial flutter ablation in addition to pulmonary vein isolation. Mean procedural time for the initial RFA was 205 min, with a mean fluoroscopy time of 49 min. The mean procedural time for repeat RFA was 146 min with a mean fluoroscopy time of 40 min. Every patient had electrical isolation of all four pulmonary veins, with an additional adjoining roof line and further ablation at sites of complex fractionated electrograms. One patient reverted to atrial flutter and after flutter ablation, cardioverted to sinus rhythm at the end of the procedure. All other patients remained in AF at the end of the procedure and required internal electrical cardioversion to restore sinus rhythm.

Maintenance of sinus rhythm

All patients in the medical treatment arm remained in AF at study completion. Of the 20 patients randomised to RFA, 10 were in sinus rhythm 3 months after the initial procedure. Two of these subsequently reverted to AF before their final study visit. Among the 10 patients who had reverted to AF by their 3-month visit, four declined a repeat procedure. Of the six patients who underwent a second procedure, only two remained in sinus rhythm at their final visit. Overall, the rate of maintenance of sinus rhythm at the end of follow-up was 10/20 (50%).

Primary end point: CMR ejection fraction

Change in LVEF (measured by CMR) from baseline to end of study was +4.5±11.1% in the RFA group and +2.8± 6.7% in the medical treatment group (p=0.6; table 2).

Table 2

Change in the primary and secondary end points over the study period

Secondary end point: radionuclide ejection fraction

The increase in radionuclide LVEF from baseline to the end of the study was 8.2±12.0% in the RFA group and 1.4±5.9% in the medical treatment group (p=0.032; table 2). There was also a trend towards a greater increase in right ventricular ejection fraction in the RFA group (+4.3±8.1% vs +1.2±4.9%; p=0.17).

Other secondary end points: 6 min walk distance, NT-proBNP and quality of life

There was no difference between treatments for change in NT-proBNP or 6 min walk distance (table 2). Change in quality of life as determined by the KCCQ and MLHFQ questionnaires did not differ between treatment groups (table 2). However, the SF-36 physical component summary score increased more in the RFA group than in the medical treatment group. (+4.0±9.5 units vs −1.0±4.4 units; p=0.042).

Other CMR indices

There was no difference between the treatment groups in change from baseline in either LVEDV or LVESV. There was a trend towards reduction in LAEDA in the RFA group compared with the medical treatment group (−237±490 mm2 vs +36±364 mm2; p=0.06).

Post hoc analysis of outcomes according to achieved rhythm

We performed a post hoc analysis to compare patients in sinus rhythm with those in AF at the end of the study (tables 3 and 4). Patients in sinus rhythm were younger, less likely to have coronary heart disease and more likely to have AF of a shorter duration. Patients in sinus rhythm at the end of the study had a significantly greater improvement from baseline in CMR ejection fraction (+10.4±10.4%) than those in AF (+1.5±7.7%; p=0.008). Most of this improvement was apparent within 1 week of RFA. There was also an increase in radionuclide left (+13.2±15.0% vs +2.1±5.6%; p=0.045) and right ventricular ejection fraction (+8.5±9.2% vs +0.8±4.5%; p=0.029) and a reduction in LAEDA (−389±377 mm2 vs −22±450 mm2; p=0.034) in patients in sinus rhythm compared with those in AF. There was no difference in change in quality of life or NT-proBNP between those in sinus rhythm and those in AF at the end of the study.

Table 3

Atrial fibrillation (AF) v sinus rhythm (SR) at study completion: baseline characteristics

Table 4

AF v SR at study completion: end points


One patient had a stroke 6 days after their initial procedure and withdrew from further follow-up. Two patients developed cardiac tamponade during RFA. Both recovered after emergency pericardiocentesis without long-term complications. Three further patients developed worsening heart failure: the first required intravenous diuretic immediately following the procedure; the second required an increase in dose of oral diuretic before discharge; the third required admission 2 days after the procedure for 3 days of intravenous diuretic treatment. No pulmonary vein stenosis was identified. Overall, four major complications (cerebrovascular accident, cardiac tamponade or readmission to hospital within 1 week) occurred in relation to a total of 27 procedures, giving an overall major complication rate of 14.8%.


In patients with advanced heart failure, severe left ventricular dysfunction and persistent AF, we found that a strategy of rhythm control using RFA, compared with a strategy of rate control using medical treatment, did not improve our primary outcome of LVEF, measured using CMR. However, patients randomised to RFA had a greater increase in radionuclide ejection fraction (a prespecified secondary outcome measure) than those assigned to medical treatment alone. There were no significant differences in the other prespecified secondary outcome measures, including changes in cardiac structure, NT-proBNP and 6 min walk distance. There was also no consistent improvement in quality of life after RFA compared with medical treatment.

It is uncertain how to interpret the discrepancy between the CMR and radionuclide ejection fraction measures. The strictest interpretation is that because CMR was the primary outcome measure, little value should be place upon the radionuclide findings. However, radionuclide ventriculography is a well-established and reliable technique for measuring ejection fraction and may even be more reliable than CMR at measuring ejection fraction in patients with AF. With radionuclide ventriculography, LVEF is measured on a composite beat created from 20 min of recorded heart beats, whereas CMR cine images are a composite of the cardiac cycles that can be acquired during one breath-hold.

Our ability to demonstrate an improvement in LVEF may also have been reduced by the low rate of maintenance of sinus rhythm after RFA in this trial. Most other studies of RFA in patients with left ventricular systolic dysfunction have reported higher success rates than ours—between 50% and 78% after a first procedure and between 78% and 96% after a second procedure. Our lower success rate was probably explained by differences between the patients in the current trial and those in previous studies (table 5). Our patients were older, had more severe systolic dysfunction and all had longer-standing and persistent AF. These clinical characteristics promote atrial dilatation and fibrosis, which are both independent predictors of decreased procedural success rate after RFA.19 20 Previous studies mainly recruited younger patients referred for RFA because of symptomatic paroxysmal or persistent AF and incidentally found to have reduced left ventricular systolic function.11–14

Table 5

Previous studies examining radiofrequency ablation (RFA) for atrial fibrillation (AF) in patients with LVSD/HF

To take account of the lower than expected rate of successful restoration of sinus rhythm, we examined change in cardiac structure and function according to the rhythm achieved. In this post hoc analysis we saw marked improvements in those who remained in sinus rhythm following RFA compared with those who remained in AF at the end of the study. These improvements included an increase in LVEF (measured by CMR and radionuclide ventriculography), as well as improved right ventricular ejection fraction and reduced atrial size. Interestingly, in this analysis, an increase in LVEF (measured by CMR) was measurable within 1 week of RFA, suggesting that restoration of sinus rhythm, rather than ventricular remodelling, was responsible. However, this early change also raises the possibility that the observed change in ejection fraction may simply be a consequence of measuring ejection fraction in sinus rhythm rather than in AF. Thus, whether the apparent increase in ejection fraction truly represents improved cardiac function or is merely a measurement issue is unclear. In this respect, it is notable that patients with sustained sinus rhythm did not show a significant improvement in NT-proBNP, 6 min walk distance or quality of life, although we had very limited power to detect anything other than a large treatment effect.

It is difficult to compare our trial with previous studies which examined the effect of RFA on left ventricular function in patients with heart failure and AF. Each used echocardiography to assess ventricular function and most (but not all) reported that RFA increased ejection fraction. However, with one exception, all of these were non-randomised and observational, making interpretation of their findings difficult. The only other randomised controlled trial of RFA in patients with heart failure and AF, the Pulmonary Vein Atrium Isolation versus AV Node Ablation with Bi-Ventricular Pacing for Treatment of Atrial Fibrillation in Patients with Congestive Heart Failure study (PABA-CHF), had a quite different design than our trial. In PABA-CHF, patients with paroxysmal or persistent AF were randomised to either RFA (n=41) or atrioventricular node ablation together with implantation of a cardiac resynchronisation-defibrillator device (n=40).18 Ejection fraction was measured using echocardiography. Change in ejection fraction from baseline was higher in the RFA group (+8.8±8%) than in the atrioventricular node ablation group (−1±4%) after 6 months (p<0.001). Patients in the RFA group also exhibited greater improvement in quality of life and functional capacity than the other group.

In addition to the method of measuring ejection fraction, there were other notable differences between PABA-CHF and our study, including the type of AF, the age of the patients, the comparator treatment and the higher procedural success rate (88% of patients randomised to RFA in PABA-CHF were in sinus rhythm at 6 months).

In addition to the uncertain effect on ventricular function and low procedural success rate, the value of RFA in patients with advanced heart failure and persistent AF is further questioned by the lack of effect on other outcomes (NT-proBNP, quality of life and 6 min walk distance) and, particularly, the relatively high complication rate associated with this technique in such patients. Even though all procedures were performed by a single experienced operator, one patient had a stroke and two experienced cardiac tamponade. Previous studies of AF ablation reported lower rates of these serious complications. However, the risk of tamponade may relate to the larger atria of older patients with advanced heart failure who are also at higher risk of stroke.

Our trial has a number of limitations. Our sample size was based on a predicted success rate of 80% in restoring long-term sinus rhythm. The lower than predicted success rate increased the risk of a type II statistical error—that is, of missing a real difference. Despite randomisation, there were some baseline imbalances between the treatment groups. Ours was a single-centre study with a single operator performing RFA.

In conclusion, RFA resulted in long-term restoration of sinus rhythm in only 50% of patients with symptomatic systolic heart failure and persistent AF. Analysed by intention to treat, a strategy of rhythm control using RFA did not improve the primary outcome of LVEF measured by CMR compared with a strategy of rate control with medical treatment. RFA did improve ejection fraction measured using radionuclide ventriculography but did not improve other secondary outcomes and was associated with a significant rate of serious complications. Because a post hoc analysis suggested that achievement of sinus rhythm was associated with improved cardiac structure and function, future studies may be considered in patients with a greater chance of procedural success and lower risk of complications.


We would like to thank the patients for their participation.


View Abstract


  • See Editorial, p 687

  • Institution where work was performed: Golden Jubilee National Hospital, Glasgow.

  • Funding This work was supported by the Chief Scientist Office, Scotland (grant number CZB4475).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the West Glasgow REC.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles