Late outcome of surgical radiofrequency ablation for persistent valvular atrial fibrillation in China: a single-center study

We consecutively recruited 510 valvular AF patients with written informed consent admitted to the Affiliated Drum Tower Hospital of Nanjing University Medical School between April 2010 and November 2014. A pre-ablation 7-day Holter and comprehensive transthoracic echocardiographic examination were performed to establish that all patients had persistent valvular AF. The study was conducted according to the Helsinki Declaration and approved by the ethics committee of Nanjing University.

Complete clinical data were collected before discharge. There were 114 patients excluded: 11 cases died in hospital due to low cardiac output syndrome and multiple organ failure; 2 cases died of sudden death after discharge; 2 case died of heart failure after discharge; 8 cases died of trauma or cerebrovascular accident; 7 cases had a history of implanted pacemaker before surgery; 8 cases received pacemaker implantation because of sinus long-intermission rhythm and advanced atrioventricular block. And another 76 cases were lost at follow-up. Therefore, a total of 396 patients had complete follow-ups and clinical information.

All surgical procedures were performed in standard fashion by the same surgical team through median sternotomy. All patients underwent either a mitral valve repair or replacement by cardiopulmonary bypass with a mild hypothermia (32-34 °C), antegrade cold blood cardioplegic arrest, and local hypothermia with ice-slash; cardioplegic solution was readministered every 20 to 30 min. Radiofrequency ablation was performed using a mono-polar radiofrequency device (Cardioblate®; Medtronic, USA). Left and right atrial appendages and the ligament of Marshall were incised routinely. We did vertical right atrial incision between the superior and inferior vena cava in all patients. Maze IV was performed before the other intra-cardaic procedures. The lines performed included isolation of the right and left pulmonary veins firstly and the connective lines between them. Next lines were from isolating line of the left pulmonary vein to the base of left atrial appendage amputation site and to the posterior mitral valve annulus. In the right atrium, the isthmus ablation runs from the inferior caval vein, across the interatrial septum, up to the caudal aspect of the coronary sinus ostium and over to the posterior tricuspid valve annulus. If patients had a huge left atrium, we would add several auxiliary ablation lines across left atrial roof and posterior wall. Epicardial temporary pacing wires were implanted in all patients.

Heart rhythm was continuously monitored after surgery. Dual-chamber stimulation at 80 bpm through epicardial temporary pacing wires was used in most of patients for the first 48 h post-operatively to avoid severe brady arrhythmias. After that, all patients received amiodarone and metoprolol to stable rhythm except these patients whose heart rate is too slow to tolerate these antiarrhythmic drugs. AF recurrences after surgery were treated by resuming amiodarone infusion and correcting, in the case, electrolyte abnormalities. In patients in whom AF persisted despite optimal medical therapy, at least one attempt of electronic cardioversion was performed before discharge.

Late recurrence was defined as any episode of AF, atrial flutter or atrial tachycardia that lasted greater than 30 s at least 6 months after surgery [8]. AF-free time was calculated from the date of ablation to the date of recurrence or last follow-up 28 (6 ~ 61) months. Patients had scheduled clinical visits. 12-lead electrocardiography and 24-h Holter monitoring were monthly conducted after 6 months within the first year and then chosen every half year after that. Moreover, patients would receive electrocardiography monitoring in local clinics at anytime if they had AF-related symptoms. Within the first 6 months, patients who experienced AF recurrence would receive electronic cardioversion to maintain SR. Late recurrence of AF during the follow-up was considered censored.

For the comparison between the two groups, Student’s t–test (normally distributed) or Mann–Whitney test (non-normally distributed) was used for continuous variables, and χ² test was utilized for categorical variables. Cox survival regression model was used to determine factors predictive of AF late recurrence. Factors for which the univariate analysis gave a p value ≤0.1 were included in the multivariate analysis. Hazard ratio (HR) and 95% confidence interval (CI) were calculated. Youden index (Sensitivity + Specificity-1) of receiver-operating characteristic (ROC) curve was calculated to single out the best cutoff values of AF duration, pre-operative B-type natriuretic peptide (BNP), pre-operative heart rate and left atrial diameter (LAD) predicting AF late recurrence. Sensitivity, Specificity and Accuracy were investigated by the Fisher’s exact test. Survival curves for the incidence of AF recurrence by AF duration, pre-operative BNP, pre-operative heart rate and LAD were calculated with the Kaplan-Meier method and a log-rank test was used to assess statistical significance. p < 0.05 was considered statistically significant, and all statistical tests were two-sided. The statistical analysis was performed with the GBSTAT statistical analysis package (version 9.0, Dynamic Microsystems, Inc).

The 396 patients were divided into two groups according to whether AF late recurrence took place or not (SR group n = 245, AF group n = 151, see Table 1). No significant differences were found in terms of gender, age, the ratio of rheumatic cause, cerebral infarction, pre-operative left ventricular ejection function, combined tricuspid valve annuloplasty and coronary artery bypass grafting, cardiopulmonary bypass duration, aortic clamp time, intensive care unit stay, post-operative using of amiodarone and metoprolol. The AF group had longer AF duration and higher percents of hypertension, left atrial thrombosis and redo-procedure than the SR group. However, the AF group had lower ratio of diabetes and worse New York Heart Association class than the SR group. The pre-operative echocardiographic LAD was larger and pre-operative serum BNP was higher in AF than SR, but heart rate in the AF group were slower than SR group in the admission. Moreover, the AF group experienced a shorter duration of ventilation and higher ratio of AF at discharge and electronic cardioversion within 6 months than the SR group.

In the Cox univariate regression analysis, significant predictors of AF late recurrence were AF duration, hypertension, left atrial thrombosis, New York Heart Association class, pre-operative serum BNP level, pre-operative heart rate and LAD, pre-operative left ventricular ejection fraction, redo-procedure, post-operative atrial tachyarrhythmias, AF at discharge and post-operative amiodarone.

In the multivariate model, AF duration, pre-operative serum BNP level, pre-operative heart rate and LAD, post-operative atrial tachyarrhythmias and AF at discharge remained statistically significant predictors of AF late recurrence (Table 2).

According to the ROC curve analysis, the best threshold values of AF duration, pre-operative serum BNP level, pre-operative heart rate and LAD were 66.5 months (39.7% sensitivity, 89.0% specificity and 55.3% accuracy), 251 pg/ml (60.3% sensitivity, 71.4% specificity and 67.2% accuracy), 82 beats/min (75.1% sensitivity, 81.5% specificity and 75.8% accuracy) and 67.9 mm (31.8% sensitivity, 91.4% specificity and 68.7% accuracy), respectively (Table 3).

In addition, Kaplan-Meier survival estimates showed that patients with AF duration ≥66.5 months, pre-operative serum BNP level ≥ 251 pg/ml and pre-operative LAD ≥67.9 mm had higher proportions of AF late recurrence than patients with AF duration <66.5 months (Fig. 1a), pre-operative serum BNP level < 251 pg/ml (Fig. 1b) and pre-operative LAD <67.9 mm (Fig. 1d), respectively (p < 0.001). However, Patients with pre-operative heart rate > 82 bpm had a much lower proportion of AF late recurrence than patients with pre-operative heart rate ≤ 82 bpm (p < 0.001) (Fig. 1c).

Several limitations of this study require consideration. First, the nonrandomized design may have affected the results because of unmeasured confounders, procedure bias or detection bias, despite the use of rigorous statistical adjustment. Second, because of the initial deficiency of mono-polar radiofrequency device, defined as the lack of intra-operative lesion integrity assessment, it may have left some patients with gap lesions, ultimately affecting outcome. Third, all procedures were performed in a single institution. A related potential bias may result from the reliance on self reporting. So a multi-center study with a large population is warrant to confirm our findings. Moreover, observed freedom from AF varies with the rigor of rhythm assessment. As compared with continuous long-term monitoring, spot electrocardiographic recordings tend to overestimate success by approximately 12 percentage points [26]. Therefore, a study using an implantable continuous monitor is warranted to improve our finding.