This study was a single-center, retrospective, observational study. Consecutive patients who underwent total thoracoscopic ablation for AF, from February 2012 to May 2020, were included. Patients who converted to open Cox–Maze surgery exhibited underlying moderate to severe mitral stenosis or were subjected to short-term follow-up loss after <1 year were excluded. In addition, patients who did not receive LAA occlusion were excluded. Patients who underwent unilateral pulmonary vein isolation (PVI) or thoracoscopic LAA occlusion only without AF ablation were included (Figure 1). We evaluated the development of thromboembolism during follow-ups. This study was approved by the Institutional Review Board of Samsung Medical Center, South Korea (IRB No. 2020-06-159).

Before surgery, all patients underwent transesophageal echocardiography (TEE) to exclude LAA thrombus. Total thoracoscopic ablation procedures were performed under general anesthesia. All procedures were performed using standard techniques as described previously (9). The bilateral thoracoscopic approach was used with a video-assisted thoracoscopic surgical technique. Beginning on the right side, a 5-mm port was introduced into the fourth intercostal space at the mid-axillary line. After carbon dioxide insufflation to expand the operative field and depress the diaphragm, the remaining two ports were placed into the third intercostal space at the anterior axillary line and the sixth intercostal space at the mid-axillary line. After pericardial tenting, a lighted dissector (AtriCure Lumitip Dissector, Atricure, Inc., Cincinnati, OH, USA) was used to pass a rubber band under the PV antrum through the oblique sinus. An AtriCure Isolator Transpolar Clamp (Atricure, Inc.) was connected to the rubber band and positioned around the PV antrum. PV antrum isolation was performed by applying bipolar radiofrequency energy 6 times to the clamps around the PV antrum. Additional superior and inferior ablation lines connecting both PV isolation lines were created epicardially using a linear pen device (MLP, Atricure, Inc.). Ganglionated plexi subsequently were ablated with bipolar radiofrequency energy with the aid of high-frequency stimulation. Confirmation of ablation lines was obtained by pacing testing using the AtriCure Cooltip pen (MAX5, Atricure). The procedure was repeated on the left side. Before PV and ganglionated plexus ablation, the ligament of Marshall was dissected and ablated. When all ablations were complete and the conduction block was confirmed, the left atrial appendage was removed using an Echelon Flex 60 articulating endoscopic linear stapler (Ethicon Endo-Surgery Inc., Cincinnati, OH, USA) (Figure 2).

The primary outcome was the occurrence of ischemic stroke and thromboembolism after surgery. The etiology of ischemic stroke was evaluated and determined to be procedure-related, cardio-embolic, or resulted from small artery occlusion.

All patients were followed up by 2 weeks, 3 months, 6 months, and every 6 months thereafter.

Electrocardiography (ECG) was performed at each visit, and 24-h Holter monitoring was performed at three, six, and 12 months and annually thereafter. Additional monitoring was performed when patients experienced tachyarrhythmia symptoms. Follow-up computed tomography (CT) –angiography was performed to identify residual LAA or LA thrombus formation at least 6 months afterwards. Successful LAA occlusion was defined as the absence of a residual trabeculated LAA stump. Other findings such as LA thrombus, accessory appendage, and remnant peri-occluder pouch were evaluated. Oral anticoagulant (OAC) was resumed 1 or 2 days after the procedure with complete hemostasis and continued for at least 3 months.

Statistical analysis was performed using SPSS ver. 27.0 software (SPSS Inc., Chicago, IL, USA). Continuous variables were compared using the unpaired t-test or Wilcoxon rank-sum test, and categorical variables were compared using either the Chi-squared test or Fisher's exact test as determined appropriate. The incidence rates of clinical events are presented as person-years and events rate curves were obtained by Kaplan–Meier analysis. The risk of thromboembolism was assessed using a Cox proportional hazards model, and is presented as the hazard ratio (HR). A p-value < 0.05 was considered significant.

The total number of 521 patients underwent the total thoracoscopic ablation procedure, and 61 were excluded from the study. Fifteen patients were converted to open Cox–Maze surgery, three patients were mitral stenosis, 16 patients were lost to follow-up after <1 year, and 27 patients did not receive LAA occlusion due to advanced heart failure or small LAA. 460 patients [mean age, 57.1 ± 9.2 years; 400 (87.0%) males] were included for analysis. Among these patients, 14 received only thoracoscopic LAA occlusion without AF ablation. The mean follow-up duration was 44.8 months. 385 (83.7%) patients exhibited persistent AF and 94 (20.4%) had a previous history of ischemic stroke. The median OAC duration was 109.5 days after the surgery, and the total of 411 (89.3%) patients discontinued OAC. Anticoagulation discontinuation rate according to CHA₂DS₂-VASc score are as follows; (i) 0 = 99.0%; (ii) 1 = 98.2%; and (iii) ≥2 = 81.3%. The mean CHA₂DS₂-VASc score was 1.9 ± 1.6, and 104 (22.6%) patients exhibited score 0 (Figure 3). The mean CHA₂DS₂-VASc score excluding the score 0 was 2.4 ± 1.3. Follow-up CT was performed in 337 (73.3%) patients, and 26 (7.7%) patients exhibited remnant LAA. Total 277 (60.2%) patients have maintained sinus rhythm during overall follow-up. The one- and two-year atrial tachyarrhythmia–free survival rates were 84.5 and 70.9%, respectively. The annualized recurrence rate of atrial tachyarrhythmia was 14.2%/year. The other baseline characteristics are summarized in Tables 1, 2.

The annualized incidence rate of ischemic stroke was 0.78%/year, comprising 13 patients, which demonstrated a 73% risk reduction compared with the CHA₂DS₂-VASc-predicted rate without anticoagulation (10). Four patients exhibited procedure related ischemic stroke – three occurred <4 days after the surgery and one was stroke related with an atrial-esophageal fistula after 1 month of the surgery. Four patients had a history of previous embolic stroke. Only one patient experienced an ischemic stroke during OAC therapy. That patient had a history of three previous embolic strokes, and follow-up CT revealed remnant LAA (Figure 4). Patient characteristics are summarized in Table 3. The annualized incidence rate of ischemic stroke excluding the CHA₂DS₂-VASc score 0 was 0.84%/year, showing a 77% risk reduction (Figure 5). Hospitalization due to heart failure aggravation was observed in two patients during follow-up. The procedure related complications are summarized in Table 4.

The hazard ratio for ischemic stroke according to previous stroke history was 1.5 [95% confidential interval (CI) 0.3–7.3, p = 0.62], those of remnant LAA was 5.1 (1.2–20.9, p = 0.02), and CHA₂DS₂-VASc was 2.9 (1.6–−5.2, p < 0.001). The recurrence of AF or the use of OAC was not associated with ischemic stroke. The hazard ratio of the recurrence of AF was 0.7 (0.2–2.7, p = 0.57), that of OAC use was 2.6 (0.4–15.5, p = 0.30) and age was 0.9 (0.8–1.0, p = 0.21) (Figure 6).

LAA represents the sources of thrombus formation in patients with AF associated with blood stasis. Decreased LAA peak flow velocity is associated with increased thromboembolic risk (11). In addition, LAA morphology correlates with the risk of stroke, especially with an increased number of lobes independent of blood stasis (12). In this regard, several methods were used to perform occlusion of LAA including surgical resection or LAA device occlusion. LAA occlusion with a device can be considered in patients with higher stroke or bleeding risk. The ischemic stroke rate was reduced by 67% with device closure. However, device-related thrombosis was seen in 1.6% of patients (13). Recently, LAA occlusion during cardiac surgery among patients accompanied with AF exhibited a benefit for ischemic stroke prevention compared with the non-occlusion group (4). Several methods including amputation and closure, stapler closure, double-layer linear closure, or closure with a surgical occlusion device were performed without complications or increase in the risk of heart failure or major bleeding. The study included patients scheduled to undergo cardiac surgery with cardiopulmonary bypass and excluded those who underwent off-pump surgery. Cardiopulmonary bypass surgery itself has the risk of thrombus formation compared with off-pump surgery (14). Therefore, early events during the first 30 days after the surgery showed no difference between the two groups associated with peri-procedural stroke. In addition, concomitant surgical ablation of AF was performed in about 30% of patients in both groups in the study. Another small study demonstrated that thoracoscopic ablation with appendage ligation could prevent recurrent stroke in patients with AF with a previous stroke compared to medical therapy (15). In our study, we conducted minimally invasive surgery using video-assisted thoracoscopy concomitant with AF rhythm control surgery without cardiopulmonary bypass. 94 (20.4%) patients had a previous history of ischemic stroke. Although, peri-procedural stroke was observed only in 0.8% of patients. Furthermore, 97% of patients underwent surgical ablation of AF. Maintaining sinus rhythms can have an influence on the risk reduction of thrombus formation after surgery.

The function of the LAA is not well-known. In animal studies, removal of the LAA resulted in decreased compliance of the LA, which was associated with decreased reservoir function (16). In our study, only two exhibited aggravation of heart failure during follow-up. We conducted LAA occlusion using an articulating endoscopic linear stapler targeting the removal of the trabeculated portion and preservation of the basal portion of the LAA, which was confirmed with CT angiography after surgery. This might have affected LA function maintenance with lowered risk of thrombus formation.

Most of the patients discontinued OAC during the follow-up, and the median OAC use-duration was 109.5 days in our study. Nearly 90% of patients discontinued OAC but showed 73% risk reduction compared with the CHA₂DS₂-VASc predicted rate without anticoagulation. Very few patients developed ischemic stroke after OAC discontinuation.

Lee et al. reported that LAA ligation or stapled excision may increase the embolic risk compared to surgical excision technique due to incompletely elimination of LAA (17). Our study showed that the remnant trabeculated LAA or peri-occluder pouch confirmed with CT was associated with increased ischemic stroke risk. The previous history of stroke exhibited a tendency to increase the risk of ischemic stroke. Therefore, CT findings after surgery were important in determining whether to continue anticoagulation therapy. In addition, patients who had a history of recurrent ischemic stroke should consider continuing OAC therapy.