Long-term Survival and Stroke after Cardiac Surgery with Concomitant Surgical Ablation for Atrial Fibrillation

Surgical ablation (SA) is an effective way to restore and maintain sinus rhythm in patients with AF.¹ The prevalence of AF in patients undergoing cardiac surgery varies widely but is estimated at 10–28% and rising.^2,3 Increasing evidence for the cardiovascular benefit of early rhythm control and for the efficacy of SA – including in more contemporary hybrid strategies – suggests a need to address AF at the time of cardiac surgery.^4,5 A number of large observational studies have demonstrated a reduction in mortality with concomitant SA.^6–8 Accordingly, the latest international guidelines recommend concomitant SA in patients with AF, citing stronger evidence for mitral valve (MV) or left atrial open surgery.^9,10

Despite these recommendations, SA remains underperformed outside certain centres and trends in the US have only increased marginally since 2016.¹¹ Data from European centres are scarce, but overall procedure uptake is likely to be similar or lower. Moreover, left atrial appendage (LAA) occlusion reduces incidence of ischaemic stroke but performance rates are seldom recorded in nationwide cohorts or registries reporting on SA.¹² We sought to analyse safety, efficacy and long-term mortality and stroke outcomes in our centre from >10 years of experience performing SA in a majority of patients with AF requiring valve surgery and/or coronary artery bypass grafting (CABG). LAA occlusion and other relevant factors were also accounted for.

Methods

Study Population

A retrospective observational study was performed on all consecutive patients with AF who underwent cardiac surgery between April 2011 and March 2022 at South Yorkshire Cardiothoracic Centre, Northern General Hospital, Sheffield, UK. Patients were included even if AF was first detected during the admission leading up to cardiac surgery, but not if AF developed solely post-operatively. Pre-operative demographic data was obtained from our institutional database.

Surgery was performed for a separate indication to AF, i.e. CABG, valve replacement/repair or a combination of these. Patients undergoing concomitant surgery on the aorta, septal myomectomy, atrial septal defect closure and/or left atrial mass excision were excluded. Standalone SA and/or LAA occlusions were also excluded. Patients were grouped based on whether they also received concomitant SA for AF at the time of cardiac surgery. In the SA group, ablation lesion sets included a biatrial Cox-Maze IV, incomplete left atrial only – defined as missing at least one left atrial line (e.g. superior or inferior) but including bilateral pulmonary vein isolation as a minimum, and incomplete biatrial ablation. Bipolar radiofrequency ablation and cryoablation were used in combination to deliver endocardial and epicardial lesions in the majority of cases, as recommended in clinical practice guidelines.⁹ Confirmation of electrical isolation via exit block was performed in a minority of cases. The decision whether to perform SA, and to what extent, was based on the operating surgeon’s clinical judgement before or at the time of surgery. LAA occlusion was an adjunctive procedure carried out via external resection/exclusion, endocardial obliteration or the use of a clip and was dependent on favourable anatomy. Long-term oral anticoagulation post-procedure was prescribed for 83% of patients with available data (97% of the total cohort), including the majority of those who also underwent LAA occlusion.

Clinical Outcomes

Post-operative complications were recorded from our institutional database. Long-term stroke data were collected from primary, secondary and tertiary care National Health Service records, and diagnoses were confirmed using CT/MRI scans previously obtained as part of routine clinical care. Mortality data were collected from Sheffield Teaching Hospitals National Health Service Foundation Trust Information Services and our institutional database. Rhythm outcomes were determined via 12-lead ECG, Holter monitoring and implantable cardiac device interrogation undertaken at 3, 12 months and annually thereafter for most patients in both groups, though with a gradual decline in long-term rhythm follow-up. Rhythm was assessed by ECGs at most follow-up visits, with prolonged monitoring being used in 52% (322/619) of patients who underwent SA. Arrhythmia recurrence was defined as any evidence of atrial tachyarrhythmia (ATA; AF, atrial flutter or atrial tachycardia) on one standard 12-lead ECG or lasting longer than 30 seconds on Holter or device interrogation.

Statistical Analysis

Normality was assessed using the Kolmogorov–Smirnov test. Categorical variables are presented as total numbers (proportions) and continuous variables as mean (SD) or median (interquartile range; IQR), as appropriate. Categorical variables were compared using the two-tailed χ² test, and continuous variables were compared using the unpaired Student’s t-test for normally distributed data and the Mann–Whitney U-test for data with skewed distribution (both two-tailed).

Propensity score matching was used to improve the balance in variables and reduce possible selection bias between patients undergoing surgery with or without concomitant SA. The 21 variables selected were: age, sex, BMI, type of AF (paroxysmal or non-paroxysmal), duration of AF, left atrial volume index as a measure of left atrial size, left ventricular ejection fraction, CHA2DS2-VASc score, New York Heart Association class, hypertension, diabetes, chronic lung disease, previous cardiac surgery, previous stroke, previous MI, extracardiac arteriopathy, pulmonary hypertension, latest creatinine clearance before surgery, operative urgency (elective or emergency), surgical access (sternotomy or minimally invasive) and number of procedures not including SA. After generating propensity scores for patients with complete data for the selected variables, matching was carried out between the two groups (with or without SA) via greedy nearest neighbour matching without replacement. A tolerance level of 0.1 SDs of the logit of the calculated propensity score was used. Covariate balance in the matched groups was assessed using standardised mean differences (SMD).

Binary logistic regression was used to calculate ORs for post-operative complications between matched groups. Overall survival was compared between matched groups with Kaplan–Meier curve and log-rank analysis. Multivariate Cox proportional-hazards regression was used to calculate adjusted HRs for predictors of survival. Ischaemic stroke was analysed via competing risk regression using the Fine-Gray method to calculate subdistribution HRs (sHR). All calculations were performed using IBM SPSS Statistics 29 including the R plug-in and R 4.4.0 (The R Foundation for Statistical Computing). A p-value <0.05 was considered statistically significant and a SMD <0.1 was considered well-balanced.

Results

Demographic Data

Of the 1,205 eligible patients, 51% (n=619) received concomitant SA (the SA group) and 49% (n=586) underwent cardiac surgery without ablation (the no-SA group) (Figure 1 ). Most baseline characteristics between these unmatched groups were significantly different (Supplementary Table 1 ). In the SA group, biatrial Cox-Maze IV was performed in 51%, incomplete left atrial lesions in 38% and incomplete biatrial lesions in 11% of patients. After propensity matching, 326 patients from each group were selected and the characteristics specifically included in the model were well-balanced (Table 1 ). European System for Cardiac Operative Risk Evaluation (EuroSCORE) II remained higher in the SA group than in the no-SA group (median 4.07% versus 2.41%; p<0.001). Despite propensity matching, there also remained a higher percentage of patients in the SA group who underwent MV and tricuspid surgery and a lower percentage who underwent CABG and aortic valve surgery compared with the no-SA group.

Peri-operative Outcomes

Peri-operative characteristics and post-operative length of stay in the matched groups are documented in Supplementary Table 2. Cardiopulmonary bypass time and cross-clamp time were increased in the SA group compared with the no-SA group (median 125 versus 92 minutes and 85 versus 69 minutes; both p<0.001, respectively). A higher percentage of patients underwent LAA occlusion in the SA group compared with the no-SA group (76% versus 49%; p<0.001). Patients undergoing concomitant SA had a slightly longer length of total post-operative hospital stay (median 10 versus 9 days; p=0.048) but length of stay on intensive care was similar in both groups (median 2 days). There were no differences in post-operative complications within the first 30 days between the matched groups (Table 2 ).

Long-Term Outcomes

Post-operative freedom from ATA in matched patients who received SA was 74% (152/206) at 1 year, 58% (69/119) at 3 years, 46% (32/69) at 5 years and 32% (9/28) at 8 years (Supplementary Figure 1 ). This was much higher than the rate of freedom from ATA recorded at each time point for matched patients who did not receive SA (always ≤10%). In the subgroup of matched patients undergoing MV surgery, freedom from ATA at 3 and 5 years was 60% and 46%, respectively, in those receiving concomitant SA, compared with 9% and 8% in those who did not. In the subgroup undergoing non-MV surgery, freedom from ATA was 54% and 47% at 3 and 5 years, respectively, with concomitant SA, compared with 8% and 6% without SA.

Median follow-up for cardiovascular outcomes in the matched groups was 4.54 (IQR 2.45–7.12) years. Kaplan–Meier curve analysis demonstrated a higher survival in the group receiving SA compared with the group not receiving SA (log-rank p<0.0001) (Figure 2 ). Survival probability in the SA and no-SA groups was 81% and 68% at 5 years, and 62% and 38% at 10 years, respectively. Improved 10-year survival was also demonstrated in the SA group on multivariate Cox proportional-hazards regression (adjusted HR 0.61; 95% CI [0.45–0.82]; p=0.001) (Table 3 ).

Ischaemic stroke occurred in 54 matched patients (8%) over a median follow-up of 4.33 (IQR 2.24–6.97) years. There was no difference in the rate of ischaemic stroke, with or without a 30-day post-operative blanking period, between matched patients who received SA compared with those who did not undergo the procedure (sHR 1.11; 95% CI [0.53–2.30]; p=0.790 and sHR 0.84; 95% CI [0.46–1.55]; p=0.570). Patients who underwent LAA occlusion had a lower incidence of ischaemic stroke with or without a 30-day post-operative blanking period compared with patients whose LAA was left untreated (sHR 0.35; 95% CI [0.17–0.74]; p=0.006 and sHR 0.54; 95% CI [0.30–0.99]; p=0.044, respectively; Supplementary Tables 3 and 4 ).

Discussion

In this propensity-matched retrospective observational study, patients with AF who received concomitant SA at the time of cardiac surgery had improved long-term survival compared with similar patients undergoing cardiac surgery without SA. Concomitant LAA occlusion was associated with a reduced incidence of ischaemic stroke, but concomitant SA was not. Early post-operative complications, including permanent pacemaker implantation, were similar in patients undergoing cardiac surgery with or without SA.

There is a well-established indication to restore sinus rhythm by SA in patients with symptomatic AF, but carrying out the procedure regardless of symptom profile has gained more recent interest in view of increasing evidence for improvement in cardiovascular outcomes with early rhythm control.^4,9,10 Few randomised controlled trials have investigated long-term outcomes after concomitant SA and were limited by their small sample size. Five-year follow-up of the PRAGUE-12 trial demonstrated no difference in cardiovascular death but a reduced incidence of stroke, and the Amaze trial showed no difference in mortality, stroke and quality of life at 5 years.^13,14 In a retrospective study from St Louis, US, patients undergoing concomitant SA (all with biatrial Cox-Maze IV lesion sets) had improved 10-year survival compared with patients with AF who did not receive SA at cardiac surgery and similar survival to patients undergoing cardiac surgery without a history of AF.⁶

Our study included procedures from a single high-volume centre where SA is undertaken in approximately half of all patients with AF. Like previous studies, most of the variables selected for propensity score matching were components of EuroSCORE II but, unlike other studies, we also included notable predictors of AF recurrence such as left atrial size, AF duration and AF type (paroxysmal versus non-paroxysmal) as these could have influenced the decision whether or not to perform SA.^1,15 Furthermore, they could themselves also be predictors of mortality and/or stroke.¹⁶ All relevant baseline characteristics were similar between the groups after propensity score matching, but EuroSCORE II remained higher in patients receiving SA, which likely reflects the increased ‘weighting’ attributed to it as an additional procedure. Even with these uniquely matched groups, our findings related to overall survival are consistent with previous studies.

Despite the lack of trial evidence for direct association between rhythm and mortality outcomes, SA is reported to have a high rate of long-term efficacy; this is understood to be the underlying reason for improved survival compared with untreated patients. In our study, freedom from ATA after SA was lower than the rates reported by highly experienced centres with exclusive use of Cox-Maze lesion sets, where ≥80% at 5 years and well over 50% at 7–8 years were achieved.^1,15 Some important predictors of ATA recurrence in our cohort, such as an increased median left atrial volume and AF duration, could not be compared with cohorts from other studies. Moreover, assessing rhythm outcomes in a retrospective study can be limited by the loss of patients to follow-up and the extent of monitoring with which to determine arrhythmia freedom. Nevertheless, there was a clear difference between rates of ATA freedom with and without concomitant SA at every timepoint.

Concomitant SA did not significantly reduce the long-term incidence of ischaemic stroke, but event rates were lower in patients who underwent concomitant LAA occlusion when adjusting for both procedures. These results are in keeping with the main findings from LAAOS III, suggesting no impact on stroke outcomes from SA.¹² In the PRAGUE-12 study, LAA resection was performed for all patients in the SA group and in the study by Kim et al., detailed information on LAA occlusion was not available.^8,13,17 Our study provides further evidence to suggest that LAA occlusion is likely to be the main driver behind a reduction in ischaemic stroke risk rather than SA, but future randomised controlled trials could help to discern the contribution of either procedure.

The latest Society of Thoracic Surgeons guidelines provide a clear recommendation in favour of concomitant SA for any first-time non-emergency surgery, while European guidelines provide a stronger recommendation for MV surgery.^9,10 However, recent reports estimate that concomitant SA was performed in <50% and <20% of non-emergency surgeries in the US and Poland, respectively.^11,18 The ability to overcome the various barriers that are driving persistent underperformance will differ between centres, but emphasising the safety and efficacy of concomitant SA remains an important step. Our study did not reveal any major safety concerns with the addition of SA despite the expectedly increased cardiopulmonary bypass and cross-clamp time. Importantly, we did not observe an increase in early post-operative permanent pacemaker implantation (<4% in both matched groups), although the practice in our centre is to wait longer for rhythm recovery than current recommendations advise, and event rates were within the incidence range reported in other European studies.¹⁸ Though efficacy is challenging to monitor and may drop markedly in the long term, this study demonstrated that restoration and maintenance of sinus rhythm, or even staving off AF recurrence for a few years at least, improves survival and should be pursued when the opportunity presents itself. An upcoming trial (NCT05434819) aims to further investigate the benefits of concomitant SA on cardiovascular outcomes, primarily heart failure admissions.

This study had a number of limitations. First, it was inherently limited by its retrospective and observational nature. Second, although propensity score matching aimed to balance the groups regarding comorbidities and operative risk, certain variables, such as procedure type, could not be included as the differences between them were too great. In keeping with current trends elsewhere, SA tends to be performed more often with concomitant MV surgery and tricuspid surgery than with concomitant aortic valve surgery or CABG. Nevertheless, they were adjusted for in the Cox proportional-hazards model assessing survival. Third, rhythm outcomes were limited by loss to regular follow-up as previously explained, and competing risk analysis on ATA recurrence could not be accurately performed. AF burden, which is increasingly being recognised as a fairer measure of ablation success, could not be assessed and we resorted to a stricter method for determining ATA freedom, which may not have been an accurate representation of the true cardiovascular benefit of the procedure. Fourth, we did not have complete data on duration of post-operative oral anticoagulation and antiarrhythmic drugs, but our standard practice is to prescribe amiodarone for 6–8 weeks if patients develop post-operative AF, especially if they received SA.

Conclusion

In patients with AF undergoing cardiac surgery, concomitant SA can be performed without increased risk based on the findings from this cohort. It provides acceptable rhythm control in the setting of advanced AF and is associated with improved long-term survival. Cardiac surgery provides a reasonable opportunity to treat AF with SA and LAA occlusion, which could translate into long-term cardiovascular benefits.

Click here to view Supplementary Material.