OUP user menu

Resumption of sinus rhythm in patients with heart failure and permanent atrial fibrillation undergoing cardiac resynchronization therapy: a longitudinal observational study

Maurizio Gasparini, Jonathan S. Steinberg, Aysha Arshad, François Regoli, Paola Galimberti, Arnaud Rosier, Jean Claude Daubert, Catherine Klersy, Ganesh Kamath, Christophe Leclercq
DOI: http://dx.doi.org/10.1093/eurheartj/ehp572 976-983 First published online: 12 January 2010


Aims To investigate the temporal patterns, predictors, and prognostic impact of spontaneous sinus rhythm resumption (SRR) of heart failure (HF) patients with permanent atrial fibrillation (AF) treated with cardiac resynchronization therapy (CRT).

Methods and results This multicentre, retrospective, longitudinal study analysed 330 consecutive HF patients with permanent AF treated with a CRT device (mean age 70 ± 9 years, male 83%, ischaemic aetiology 44%, NYHA class III–IV 93%, mean QRS duration 167 ± 40 ms, and mean ejection fraction 26 ± 7%). Clinical, echocardiographic, and outcome data were collected during follow-up. Thirty-four patients experienced SRR after CRT (10.3%) at a median 4-month follow-up. The strongest independent predictors were end-diastolic diameter (EDD) [hazard ratios (HR) 4.03, 95% confidence intervals (95% CI) 1.43–11.36, P = 0.008], post-CRT QRS ≤150 ms (HR 2.63, 95% CI 1.02–6.67, P = 0.05), left atrium (LA) diameter ≤50 mm (HR 4.76, 95% CI 1.72–11.82, P = 0.002), and atrioventricular junction (AVJ) ablation (HR 4.27, 95% CI 1.54–11.84, P = 0.02). The coexistence of three predictors vs. zero to two predictors increased by 3.5-fold the likelihood of SRR; while the presence of all four factors improves the probability by a factor of 5.7-fold. Sinus rhythm resumption was associated with a significantly better long-term survival (log rank P = 0.03).

Conclusion One in every 10 HF patients with permanent AF may experience SRR after CRT. Baseline EDD ≤65 mm, CRT-paced QRS ≤150 ms, LA ≤50 mm, and AVJ ablation appear to be predictive of this phenomenon.

  • CRT-D
  • AVJ ablation
  • Atrial fibrillation
  • Heart failure
  • Sinus rhythm
  • Resynchronization


In patients with heart failure (HF) and permanent atrial fibrillation (AF) who undergo cardiac resynchronization therapy (CRT), limited data suggest the possibility of spontaneous restoration of sinus rhythm (SR) during follow-up. However, only few and anecdotal case reports of SR resumption (SRR) of permanent AF after CRT have been described until now.1,2 In addition, it is unknown whether SRR may be correlated to effective reverse remodelling and whether this phenomenon may yield a better prognosis during follow-up.

The aims of this study were three-fold:

  1. to investigate the incidence, temporal pattern, and predictors of spontaneous SRR in patients with permanent AF who underwent CRT implantation;

  2. to investigate whether SRR may be correlated to reversal of maladaptive remodelling after CRT implantation;

  3. to determine whether SRR may be associated with a more favourable long-term outcome.


Study design

This multicentre, retrospective, longitudinal study was conducted by three centres in Europe and the USA.

Patient selection

All consecutive patients undergoing implant of a CRT device between September 1997 and February 2008 with permanent AF were included in the study. Permanent AF was defined as AF for more than 12 months and refractory to drugs and/or repeated electrical cardioversions. The time period of AF was established by ECG and/or 24 h Holter ECG documentation. Because no clear CRT guidelines for patients with permanent AF were outlined until recently,3 HF patients with AF who satisfied all CRT indication criteria except SR were included. All patients were on optimal medical treatment, including beta-blockers, ACE-inhibitors, diuretics, rate control medications, and aldosterone antagonists. Treatment was recorded and optimized at each clinical assessment.

Definitions and follow-up

Sinus rhythm resumption was defined as the spontaneous resumption to stable SR observed during follow-up after CRT. Only SRR confirmed at two outpatient visits at least 1 month apart was considered as a primary endpoint. Patients who converted to SR following defibrillation testing or cardioversion during the follow-up were not considered as spontaneous SRR, and were therefore excluded from the analysis. If SR was maintained for more than 1 month, an upgrade to an atrial lead was systematically performed.

For the identification of the predictors of SRR, 14 baseline variables were considered: gender, aetiology of HF, NYHA class, baseline spontaneous and post-CRT QRS duration, 6 min walk test (6MWT), pre-implant echocardiographic characteristics [ejection fraction (EF), end-diastolic diameter (EDD), end-systolic diameter, mitral regurgitation (MR)], atrioventricular junction (AVJ) ablation during the follow-up, amiodarone treatment, and the degree of reverse ventricular remodelling at 6 months, defined as decrease in end-systolic volume (ESV) ≥10%4 or increase in EF >3%, at the last visit before SRR.

Outcome data were collected at each participating centre by reviewing outpatient clinical files or by phone interviews with relatives and/or physicians. Deaths were classified as cardiac, non-cardiac, or unknown. Cardiac deaths were classified as sudden (not preceded by HF or ischaemic symptoms) or owing to HF. Patients undergoing left ventricle assist device or urgent heart transplant were censored and classified as HF deaths. When the cause of death could not be determined using all available sources, it was classified as unknown.

Statistical analysis

Data were described as mean and standard deviation (SD) or median and 25th–75th percentiles if continuous, and as counts and per cent if categorical. Median follow-up (25th–75th percentiles) was computed according to the inverse Kaplan–Meier method. For the purpose of the analysis, continuous variables were dichotomized at their median. Event rates per 100 person-years were computed together with their 95% CI. Event-free survival was computed with the Kaplan–Meier method and was compared between groups with the log-rank test. Hazard ratios and 95% CI were calculated by means of the Cox proportion hazard regression model. The proportional hazard assumption was tested, based on Schoenfeld residuals. Left ventricle remodelling and AVJ ablation were treated as time-dependent co-variates. Variables showing a P-value <0.1 derived from the univariable analysis were included in a multivariable Cox model. Model discrimination was assessed graphically by plotting the survival curves of prognostic index, categorized into its tertiles; model calibration by calculating the shrinkage coefficient. All models were stratified by the implanting centre. Logistic regression was used to assess the association of remodelling and SRR; Huber–White robust standard errors were computed to account for intra-patient correlation over visits. For assessing the association of SRR and mortality, the observation time started at the sixth month assessment date when reverse remodelling effect would be expected after CRT.5 The log-rank test was used to compare survival curves. Stata 10.1 (StataCorp, College Station, TX, USA) was used for computation. A two-sided P-value <0.05 was considered statistically significant.


Three hundred and thirty patients were enrolled in the study. They were followed up for a median of 42 months (25th–75th 19–65 months). The characteristics at the CRT implant of study patients are summarized in Table 1. They were largely elderly men, with a mild impairment of renal function; 44% had ischaemic aetiology of HF and a minority had diabetes. The lateral and postero-lateral coronary sinus branches were most frequently utilized for LV lead position; a combined CRT–ICD device was implanted in half of the population; and an upgrade from a previous PM or ICD was performed in 16% of cases. Atrioventricular junction ablation was performed in 76% of cases (251/330) to ensure permanent biventricular pacing defined by >85% of pacing.5 Most of the patients were in NYHA class III, and the mean EF was <30%. More than 80% of the cohort was treated with ACE-inhibitors, beta-blockers, and diuretics; 60% of these patients were treated with aldosterone antagonists.

View this table:
Table 1

Baseline characteristics of the study cohort

CharacteristicDescription n (%) or mean ± SD
Age (years)70 ± 9
Male gender274 (83%)
Ischaemic aetiology146 (44%)
Diabetes (no/NIDDM/IDDM)145 (82%)/23 (13%)/9 (5%)
NYHA III–IV296 (94%)
BUN (mg/mL)62.3 ± 35.6
Creatinine (mg/mL)1.49 ± 0.86
Hospitalization (previous year)3 (2–4)a
Spontaneous baseline QRS (ms)167 ± 40
6MWT (m)310±111
Implantable cardioverter defibrillator158 (48%)
Upgrade from previous PM or ICD54 (16%)
Post-CRT QRS (ms)152±25
LV region (ant/antlat/lat/postlat/middle)20 (7%)/50 (17%)/117 (39%)/99 (33%)/11 (4%)
Echocardiographic data
 Ejection fraction (%)26.4 ± 7.3
 End-diastolic diameter (mm)65.3 ± 8.9
 End-systolic diameter (mm)54.1 ± 10.6
 End-diastolic volume (mL)186.2 ± 57.6
 End-systolic volume (mL)131.8 ± 45.8
 Mitral regurgitation grades 3–452 (22%)
 Left atrium diameter (mm)50.2±8.3
Medical therapy
 Angiotensin-converting enzyme inhibitor146 (82%)
 Angiotensin receptor blockers27 (15%)
 Beta-blocker145 (82%)
 Aldosterone antagonist111 (63%)
 Diuretics154 (87%)
 Amiodarone57 (32%)
 Digoxin86 (49%)
  • NIDDM, non-insulin dependent diabetes mellitus; IDDM, insulin-dependent diabetes mellitus; 6MWT, 6 min walking test.

  • amedian (25th–75th).

Resumption of sinus rhythm: incidence, temporal pattern, and predictors

Thirty-four patients (10.3%) experienced SRR at a median time of 4.4 months from implant (25th–75th, 2–9 months), yielding an incidence of SRR of 4.5 per 100 person-years (95% CI 3.5–6.2) (Figure 1). All SRR patients successfully received an atrial lead after a mean of 2 months; no infective complications were observed in this subgroup of patients; however, five patients developed a pocket haematoma, and two patients had an occlusion of the left subclavian vein. In these last two cases, the atrial lead was implanted from the right subclavian vein and tunnelled to the left pocket. Table 2 summarizes the potential predictors of SRR. Smaller EDD, shorter post-CRT QRS, smaller left atrium (LA) diameter, as well as the performance of AVJ ablation significantly increased the likelihood of SRR (approximately three- to four-fold) at univariable analysis. These same factors retained statistical significance as independent of SRR in the multivariable model (P < 0.001).

Figure 1

Kaplan–Meier estimation of cumulative SR resumption. Patients at risk are reported below the curve.

View this table:
Table 2

Predictors of sinus rhythm resumption with Cox regression

CharacteristicSR resumptionUnivariable analysisMultivariable analysis model P < 0.001
n (%)SRR rate (95% CI) (per 100 person-years)HR (95% CI)P-valueHR (95% CI)P-value
Age (years)
 ≤7122 (13)5.4 (3.6–8.2)ref.0.49
 >7112 (7)3.4 (1.9–5.9)0.78 (0.38–1.59)
 Female8 (14)5.5 (2.7–10.9)ref.0.28
 Male26 (9)4.2 (2.9–6.2)0.63 (0.28–1.40)
Ischaemic aetiology
 No17 (9)3.9 (2.4–6.3)ref.0.31
 Yes17 (12)5.1 (3.2–8.3)1.42 (0.42–2.79)
II2 (11)4.9 (1.2–19.6)ref.0.69
III–IV32 (11)4.5 (3.2–6.4)1.33 (0.32–5.54)
Spontaneous QRS
 ≤17022 (14)6.5 (4.3–9.9)ref.0.14
 >1709 (7)2.6 (1.4–5.0)0.56 (0.25–1.24)
 ≤15021 (20)8.4 (5.5–12.9)3.45 (1.39–8.33)0.0042.63 (1.02–7.15)0.046
 >1506 (6)2.0 (0.9–8.4)ref.ref.
 ≤33011 (11)3.8 (2.1–6.8)ref.0.49
 >33011 (14)4.7 (2.6–8.4)1.35 (0.58–3.11)
Echocardiographic data
Ejection fraction
  ≤ 2610 (6)2.7 (1.4–5.0)ref.0.42
  > 2624 (15)6.2 (4.1–9.2)1.35 (0.63–2.90)
End-diastolic diameter
 ≤6527 (20)9.8 (6.7–14.3)4.00 (1.75–9.1)<0.0014.03 (1.43–11.36)0.008
 >657 (6)2.3 (1.1–4.9)ref.ref.
End-systolic diameter
 ≤5325 (27)9.9 (6.7–14.6)ref.0.003
 >538 (8)3.6 (1.8–7.2)0.33 (0.15–0.73)
Mitral regurgitation
 0–224 (13)5.3 (3.6–7.9)ref.0.47
 3–47 (14)6.8 (3.2–14.2)1.38 (0.59–3.25)
Left atrium
 ≤5022 (18)8.8 (5.8–13.4)2.86 (1.20–6.67)0.0114.76 (1.72–11.82)0.002
  > 507 (8)3.0 (1.4–6.3)ref.ref.
Remodelling at 6 monthsa
 No2 (3)0.7 (0.2–2.8)ref.0.36
 Yes10 (8)3.0 (1.6–5.5)1.95 (0.42–9.00)
AVJ ablationa
 No7 (9)3.3 (1.6–7.0)ref.0.003ref.0.005
 Yes27 (11)5.0 (3.4–7.2)3.72 (1.48–9.38)4.27 (1.54–11.84)
 No14 (12)4.6 (2.7–7.8)ref.0.08ref.0.234
 Yes14 (24)11.2 (6.6–18.9)1.94 (0.92–4.09)1.62 (0.73–3.57)
  • aTime-dependent covariate. ref., reference.

We further evaluated how the co-existence of one or more of the significant predictors, specifically EDD ≤65 mm, post-CRT QRS ≤150 ms, LA diameter ≤50 mm, and AVJ ablation, would increase the likelihood of SRR. As shown in Figure 2, the presence of three favourable predictors increased the likelihood of SRR by a factor of 3.5 (HR 3.5, 95% CI 1.52–8.13, P = 0.003) with respect to zero to two predictors; the presence of all four predictors further increased the probability by 5.7 (HR 5.7, 95% CI 2.29–14.02, P = 0.000) with respect to three predictors.

Figure 2

Kaplan–Meier estimation of cumulative SR resumption according to the number of favourable factors present (derived from the multivariable analysis). Cox model P < 0.001. Three vs. zero to two predictors HR = 3.5 for SRR, four vs. three predictors HR=5.7 for SRR.

Sinus rhythm resumption and reverse remodelling

To assess reverse remodelling during follow-up, 260 patients with complete echocardiographic data were analysed. There were no differences in baseline characteristics between this subgroup and the entire patient cohort. We observed an enhanced likelihood of SRR in the presence of reverse remodelling: in fact, only 5% of patients without reverse remodelling experienced SRR, while 13% of patients who exhibited LV reverse remodelling experienced SRR (P = 0.03). Owing to the low number of events, however, this association did not reach statistical significance either in cross-sectional analysis pooling each visit in a logistic model for repeated measures (OR 3.12 95% CI 0.82–11.91, P = 0.10) or when considering remodelling at 6 months as a predictor of SRR in the time-dependent Cox regression (Table 2) with a two-fold increase in the likelihood of SRR (HR = 1.95, 95% CI 0.42–9.00, P = 0.36).

Sinus rhythm resumption and mortality during follow-up

One hundred and fifty-three patients died at a mortality rate of 19.0 per 100 person-years (95% CI 16.2–22.3); 20 died from non-cardiac causes and the remaining 133 from cardiac causes (11 sudden deaths, 113 deaths from HF, six urgent transplant, and three LVAD). Sixty-six patients were excluded from the analysis since they did not reach the 6 month follow-up: either these patients died within the first 6 months (42 patients) or the implant procedure was recent (24 patients). Therefore, 264 patients, alive and under observation at 6 months, were further followed up for a median of 40 months (25th–75th, 18–62). Fourteen (5%) resumed SR and none of these patients died, while 111 patients (44%) in the group without SRR died during follow-up (Figure 3, upper panel). Computed yearly all-cause mortality rate was therefore 18 deaths per 100 person-years (95% CI 15–21; log-rank test P = 0.029). Ninety-eight of these events were cardiac deaths (Figure 3, lower panel), with an estimated mortality of 16 cardiac deaths per 100 person-years (95% CI 13–19; log-rank test P = 0.04).

Figure 3

Kaplan–Meier analysis of cardiac survival stratified according to resumption to sinus rhythm and no resumption. Include patients with a minimum 6-month follow-up after CRT. Months of follow-up are computed beginning at the sixth month assessment. PeAF, permanent atrial fibrillation at 6 months; SR resume, resumption of sinus rhythm at 6 months.


The prevalence of AF in patients with advanced HF is reported to be as high as 50%.6 Data from the Framingham study showed that the relative risk for patients with pre-existing HF for developing AF is 4.5 in men and 5.9 in women; AF is associated with an approximate doubling in mortality in subjects (male and female) with or without pre-existing cardiovascular disease.7 Moreover, the combination of AF with ventricular conduction delay is particularly associated with cardiovascular morbidity and mortality.8 Another aspect relates to how AF in HF should be managed, since trials that have compared rhythm vs. rate control strategies in this patient population have yielded neutral outcomes. Atrial fibrillation in HF therefore remains an important epidemiological and clinical problem that warrants further investigation.9 Cardiac resynchronization therapy has been increasingly employed to treat advanced HF patients who have EF ≤35% and QRS ≥120 ms.3,10 In particular, in patients with HF, CRT may improve LV haemodynamics, while maintenance of SR may be favoured from the resulting improvement in the atrial haemodynamic milieu.11 In addition, CRT may positively affect atrial electrophysiology (especially via lessening of stretch and change in local and systemic hormonal state).12,13

Analyses from small retrospective data raise the possibility that CRT may contribute to the maintenance of SR. Hügl et al.14 reviewed AF trend data in CRT as measured by the implanted device and found that of 12 patients with permanent AF after 1 month of CRT, only five (42%) patients continued to have recorded episodes of AF after 3 months. However, the scant literature on this topic is widely variable, with one report indicating that spontaneous conversion may be 7%, while another report suggests that the spontaneous conversion rate may be as high as 75% with an antiarrhythmic drug regimen and cardioversion utilized to aggressively maintain SR.1,2 Other related data in the HF cohort with permanent AF suggests that there is no discernable clinical improvement unless AVJ ablation is performed in conjunction with CRT implantation. This hybrid approach has shown to confer improvement in terms of LV reverse remodelling4,15 as well as long-term survival, particularly through a reduction in HF deaths,16 but its effect on SRR is unknown.

Sinus rhythm resumption in permanent atrial fibrillation after cardiac resynchronization therapy

By studying SRR after CRT in patients with permanent AF, this paper sheds light into the complex, intricate relationship characterizing atrial rhythm, reverse remodelling, and outcome in these patients. To make sense of this surprising phenomenon of SRR of permanent AF after CRT, which has, until now, only been described in a limited form,1,2 this investigation has gathered the experiences from three high-volume centres. This represents, therefore, the largest cohort of HF patients with permanent AF treated with CRT and the longest follow-up. The relevance of these efforts becomes clear when one considers that HF patients with AF constitute 25% of patients implanted with CRT.17

Our study found that SR resumption occurs mostly within the first 6 months after implant of a CRT device with an incidence of around 10%. Indeed, SRR largely developed before the end of the 1st year of follow-up (median of 4.4 months). A small number of patients with SRR were also observed after many years of CRT. Multivariable analysis has demonstrated that SR resumption is associated with four easily identifiable baseline predictors, namely post-CRT QRS ≤150 ms, baseline EDD ≤65 mm, LA diameter ≤50 mm, and AVJ ablation. A shorter post-CRT QRS has already been correlated to greater ventricular electrical reversal remodelling that, theoretically, may have favourable repercussions on atrial haemodynamic and electrical activity. Furthermore, it is well established that CRT may be associated with an improvement in LA function, a reduction in LA dimension, and a decrease in MR.1,18 Although the data in the literature correlating narrow post-CRT QRS and a reversal of maladaptive remodelling are few, some studies have reported that QRS narrowing after CRT may predict a marked response after CRT. Gasparini et al.19 identified a shorter QRS after CRT as a predictor of HF remission after CRT in a univariable model.

A lower degree of baseline LV remodelling expressed through LV diameters and volumes has already been described as predictive of superior response after CRT. In this regard, Gasparini et al.19 found that HF patients who at baseline had lower ventricular volumes (LVESV <180 mL) were most likely to experience HF remission after CRT.

The correlation between smaller LA dimensions and a higher possibility of SRR during the follow-up was not surprising. Small LA diameters have been correlated to a higher success rate after electrical cardioversion20 and, more recently, have confirmed to be probably the most important pathophysiological determinant for SR conversion after CRT in permanent AF.1

Concerning AVJ ablation in patients with permanent AF treated with CRT, the need for an aggressive rate control therapy to optimize CRT in these patients has been emphasized in the recent guidelines.3 It is most probable that because this approach allows for full and complete CRT delivery, reversal of maladaptive remodelling is more likely to occur.4 Evidence derived from large observational studies has described that AVJ ablation associated with CRT confers greater increases in LVEF, greater reductions in LV geometry, greater response rates,4 and better long-term survival16 compared with AF patients with preserved AV conduction treated with CRT.

Our study observed that most patients who recovered SR were also CRT responders who experienced LV reverse remodelling. This correlation was significant in categorical terms: a significant greater number of patients experienced SRR associated with LV reverse remodelling compared with patients who achieved SRR without reverse remodelling (P = 0.03). However, owing to the low number of events, when pooling each visit in a logistic model for repeated measures, only a trend (P = 0.10) was observed despite its potential clinical relevance (OR 3.12). Further studies would be useful to substantiate these findings.

The prognostic implications of SRR were profound, and the phenomenon was found to be clearly associated with very favourable long-term survival prognosis. It is noteworthy that the spontaneous development of SR after longstanding AF was associated with a favourable outcome, in contradistinction to pharmacological efforts21 in similar populations. It is uncertain whether SR itself, and its improved atrial mechanical and electrical function, is responsible for the better prognosis, or SR is a marker of overall improved cardiac performance.

Clinical implications

It is important to stress that patients who experience SRR and who do not have an atrial lead may require an up-grading procedure to position an atrial lead. This may expose the patient to immediate or delayed surgical or infective risks.22 Israel and Barold23 previously raised the possibility to implant an atrial lead in any AF patient with CRT. On the other hand, implanting all patients with an atrial lead may imply an unnecessary increase in lead burden. Therefore, the ability to identify which HF patients with permanent AF are likely to experience this phenomenon during the follow-up is clearly important to optimally manage these patients. This study has identified specific baseline clinical characteristics that correlate significantly with SR recovery after CRT. Patients who had all four predictors had a 60% chance of SRR and those with three characteristics had a 25% chance of SRR. It may therefore be reasonable to position an atrial lead at the time of implant in these patients. Given the lack of true prospective data in this respect, in patients with three or four of these factors who do not revert spontaneously to SR within 6–12 months after CRT, it may be reasonable to adopt a rhythm control strategy that may, in some cases, effectively maintain stable SR. Clearly, there is a need for prospective clinical trials24 in this area to further explore CRT in permanent AF for prevention of future cardiovascular events and better patient management. Contextually, such studies should shed light on how reverse mechanical and electrical remodelling of the atrium may accompany that of the left ventricle after CRT, thus yielding a specific ‘predictable’ clinical effect (SRR or AF).

Study limitations

Complete echocardiographic follow-up data were available in only 260 patients; therefore, the analysis concerning reverse remodelling is limited to this subgroup of patients. However, there were no differences in baseline characteristics between this subgroup and the entire patient cohort. The performance of echocardiography during follow-up was left to the discretion of the cardiology team following the patient but to our knowledge was not influenced by factors related to AF or SRR.

This is a large retrospective longitudinal study that covers a very long follow-up time span. Given such an extensive time span and the unclear CRT indication in AF patients, it is likely that the patients implanted early were particularly compromised and presented an underlying bradycardia indication necessitating CRT implant, while patients enrolled in the most recent periods were more likely to receive an ICD backup (thus influencing mortality rates and the specifically, sudden death rates). CRT guidelines have evolved,3 and device technology as well as implantation technique have improved. The analysis performed could not adjust for all of these factors.


In this large cohort, one in every 10 patients with HF and permanent AF treated with CRT spontaneously reverted to SR during follow-up, usually soon after implant. Independent predictors of this phenomenon were found to be post-CRT QRS ≤150 ms, baseline EDD ≤65 mm, LA diameter ≤50 mm, and ablated AVJ. Additional prospective clinical studies are needed to, on one hand, substantiate these findings and, on the other hand, to improve our understanding of CRT effects on atrial structure and function.


This study was supported by IRCCS Istituto Clinico Humanitas.

Conflict of interest: none declared.


View Abstract