Totally Thoracoscopic Atrial Fibrillation Ablation With the Gemini S Bipolar Clamp: A Single‐Center Descriptive Cohort Study of 1‐Year Rhythm, Safety, and Quality‐of‐Life Outcomes

ABSTRACT

Background

Outcome data for the Gemini S bipolar radiofrequency clamp are almost absent; the only indexed report being a three‐patient case series. We describe 1‐year outcomes in a consecutive cohort.

Methods

We retrospectively analyzed all consecutive patients treated with the Gemini S clamp between January 2021 and August 2024. Epicardial box isolation was combined with left atrial appendage (LAA) exclusion; completeness was confirmed by entry and exit block on epicardial three‐dimensional mapping. The primary outcome was freedom from AF recurrence at 12 months, on and off antiarrhythmic drugs (AADs) after a 90‐day blanking period.

Results

Twenty‐three patients were included (mean age 64.4 ± 8.0 years; 82.6% male). AF was paroxysmal in 34.8%, persistent in 26.1%, and long‐standing persistent in 39.1%; 60.9% had prior catheter ablation and 69.6% had failed or were intolerant to a Class I or III agent. A complete block‐verified box lesion was achieved in 95.7% and LAA exclusion in 91.3%. Three patients (13.0%) required conversion to sternotomy for bleeding, exceeding larger series. There was no death, stroke, phrenic nerve palsy, or pacemaker implantation. At 1 year, 87.0% (95% CI 67.9–95.5) were in sinus rhythm, all off Class I/III AADs.

Conclusions

Sinus rhythm was achieved in 87.0% of patients at 1 year. As the largest reported experience with this device, the series provides device‐level evidence where little existed, but it is descriptive, not powered for comparison, and the conversion rate needs confirmation.

Thoracoscopic Gemini S box ablation with confirmed entry and exit block achieved sinus rhythm in 87.0% of 23 patients at 1 year, all off Class I/III antiarrhythmic drugs. The box remained isolated at two of three reinterventions, with recurrence outside it involving the superior vena cava in all three.

1. Introduction

Atrial fibrillation (AF) is the most prevalent sustained cardiac arrhythmia, affecting more than 37 million people worldwide and carrying substantial risk of stroke, heart failure, and premature death [1]. Antiarrhythmic drug (AAD) therapy remains the cornerstone of initial management but is of limited efficacy in symptomatic patients, and the long‐term success of catheter ablation in non‐paroxysmal AF remains suboptimal [2]. Stand‐alone and hybrid thoracoscopic ablation have therefore gained ground: hybrid ablation confers better arrhythmia‐free survival than catheter ablation in persistent and long‐standing persistent AF [3], and longer‐term follow‐up suggests durable efficacy approaching 80% [4].

Gaps nonetheless remain regarding lesion sets, patient selection, and the device‐specific performance of the tools used. Bipolar clamps differ in energy delivery, jaw geometry, irrigation, and lesion architecture, and these differences may influence transmurality and therefore durability. One such device is the Gemini S irrigated bipolar radiofrequency clamp, which creates an overlapping bipolar box lesion around the posterior left atrium.

Totally thoracoscopic ablation is itself an established procedure, and this study does not seek to re‐establish its validity. Its aim is narrower: to describe the 1‐year outcomes of a consecutive single‐center cohort treated with one specific device, for which published outcome data are almost entirely lacking. A PubMed search identified only a single indexed report specific to this clamp, a three‐patient case series in which the box lesion was shown to persist at 2 years [5]. Because device‐level outcome data are not interchangeable with technique‐level data, we regard a consecutive real‐world series as a useful addition even where it cannot support comparative inference. The primary endpoint was freedom from AF recurrence, assessed both on and off antiarrhythmic therapy; secondary outcomes were perioperative complications, mortality, and quality of life. Given the sample size, this study is descriptive and hypothesis‐generating by design: it is not powered for comparison against catheter ablation, nor for stratified analysis by AF subtype, and no such comparisons are attempted.

2. Methods

2.1. Patient Selection

We included all consecutive patients who underwent totally thoracoscopic AF ablation using the Medtronic Gemini S bipolar clamp between January 2021 and August 2024 at the City Hospital Zurich Triemli. Eligibility was confirmed through multidisciplinary review. Patients undergoing concomitant cardiac surgery or ablation with a different device were excluded. Only patients who had signed the institutional general consent in compliance with the Human Research Act (HRA) and Human Research Ordinance (HRO), as approved by the local ethics committee (BASEC 2024‐02245), were included. This study was conducted retrospectively under the general consent framework of the institution pursuant to Swiss HRA/HRO regulations.

Referral for surgical rhythm control followed multidisciplinary assessment of symptomatic AF and was not conditional on a fixed number of prior drug trials. Preoperative antiarrhythmic exposure was therefore recorded and is reported as observed, using the definition of the 2017 HRS/EHRA/ECAS/APHRS/SOLAECE expert consensus statement, in which refractoriness denotes AF persisting or recurring despite treatment with at least one Class I or III AAD at the maximally tolerated dose, or documented intolerance or contraindication to such therapy [6]. Preoperative medication was retrieved from referral documentation and the institutional pharmacy record for every patient. Antiarrhythmic histories were heterogeneous: some agents were prescribed for defined periods, some taken only as required, some discontinued for intolerance.

2.2. Surgical Procedure

All procedures were performed by a dedicated surgical team (S.P.S. and M.G.) under general anesthesia with double‐lumen endotracheal intubation. The patients were positioned supine, and preoperative transesophageal echocardiography (TEE) was performed to exclude left atrial appendage (LAA) thrombi; TEE was repeated intraoperatively to confirm LAA closure after clipping.

A bilateral totally thoracoscopic approach was employed under video‐assisted thoracoscopic surgery (VATS) guidance. Carbon dioxide insufflation (8 mmHg, 6 L/min) was used to facilitate thoracic cavity visualization. Epicardial posterior left atrial box isolation was performed using an irrigated bipolar radiofrequency clamp (Cardioblate Gemini‐S; Medtronic, Minneapolis, MN, USA) applied in a standardized box lesion set. Radiofrequency energy was delivered sequentially four times with the convexity of the clamp directed toward the atrial base and four times with the convexity facing the pericardium. The same sequence was repeated on the contralateral hemithorax.

Completeness of the lesion set was verified intraoperatively by electrophysiological testing rather than by device feedback alone. Entry and exit block across the pulmonary veins and the posterior box were assessed using epicardial three‐dimensional mapping (MapsPen; Medtronic) with pacing at 10 × 30 W, and the lesion set was considered complete only when bidirectional block was demonstrated.

LAA exclusion using an AtriClip device (AtriCure Inc., West Chester, OH, USA) was intended in all patients and performed unless the procedure was abandoned before this step or intraoperative findings precluded safe dissection. Occlusion was confirmed by TEE, defined as a residual stump of < 1 cm without flow into the LAA. Oral anticoagulation was interrupted 48–72 h before surgery, restarted on the first postoperative day, and continued for 3 months; thereafter it was individualized, with discontinuation considered only where complete LAA occlusion and sustained sinus rhythm were confirmed.

2.3. Data and Outcomes

Baseline data were retrieved from the institutional database (Dendrite Clinical Systems) and medical records: demographics, comorbidities, AF characteristics, preoperative antiarrhythmic therapy, echocardiographic parameters, operative details including lesion‐set completeness and its verification, and perioperative complications. AF duration before surgery was often documented only approximately, since the date of first diagnosis was frequently uncertain.

The primary outcome was freedom from AF recurrence (episodes > 30 s on any electrocardiographic recording) at 12 months, assessed both on and off AADs, after a 90‐day blanking period during which early arrhythmia recurrences were not counted as treatment failures. Arrhythmia recurrence was documented by 72‐h Holter ECG surveillance at 3, 6, and 12 months. AADs were continued for 3 months and discontinued at physician discretion based on rhythm status at follow‐up visits.

Secondary outcomes were all‐cause mortality (in‐hospital and 1 year) and procedure‐related complications including conversion to sternotomy, bleeding requiring reintervention, stroke or transient ischemic attack, phrenic nerve injury, and pacemaker implantation, together with length of stay, readmissions, and re‐interventions.

Health‐related quality of life (HRQoL) was assessed using the Atrial Fibrillation Effect on QualiTy‐of‐Life (AFEQT) questionnaire, a validated 20‐item disease‐specific patient‐reported outcome measure consisting of four domains (Symptoms, Daily Activities, Treatment Concern, and Treatment Satisfaction), with higher scores indicating better quality of life [7]. The Symptoms, Daily Activities, and Treatment Concern domains were assessed preoperatively and at the 1‐year follow‐up visit. The Treatment Satisfaction domain, which reflects patient appraisal of received therapy, was assessed at 1 year only, as it is not applicable prior to treatment. Questionnaires were administered as part of routine clinical follow‐up rather than under a research protocol, and domain scores were retrieved from the institutional clinical record. Domain scores were calculated with the standard AFEQT algorithm, 100 − [(sum of item severity scores—number of items answered) × 100/(number of items answered × 6)] [7]. Changes were interpreted against the published clinically important difference for the instrument, defined as a change of 5 points in the overall score, with subscale values of 7.1 for Symptoms and 5.1 for Activities of Daily Living [8].

2.4. Statistical Analysis

Continuous variables are summarized as mean ± standard deviation (SD); categorical variables as number and percentage (%). Confidence intervals for means were calculated using the t‐distribution and for proportions using the Wilson method. AFEQT domain scores are reported as observed means at baseline and at 1 year with the mean within‐domain change; no inferential testing of the change was performed, and the analysis is descriptive.

No a priori power calculation was performed, given the retrospective all‐comers design. A post hoc calculation illustrates the constraint: with 8 patients with paroxysmal and 15 with non‐paroxysmal AF, a two‐sided Fisher exact test at α = 0.05 has 0.4% power to detect a difference in 1‐year freedom from AF of 100% versus 80%, and approximately 80% power only if the non‐paroxysmal rate were as low as 40%. Subtype‐stratified outcomes are therefore reported descriptively, with exact confidence intervals and without significance testing, and no comparison between subtypes is implied. This study is descriptive and hypothesis‐generating; it is not powered for subgroup analysis or for comparison with catheter ablation. Analyses were performed using SAS 9.4 with the latest maintenance updates.

3. Results

3.1. Patient Characteristics and Procedural Details

A total of 23 patients were included, of whom 82.6% (19/23) were male, with a mean age of 64.4 ± 7.99 years (95% CI 60.9–67.9) (Table 1). The mean body mass index was 27.7 ± 3.99 (95% CI 26.0–29.4) kg/m², indicating a predominantly overweight cohort. Regarding AF subtype, 34.8% (8/23) presented with paroxysmal AF, 26.1% (6/23) with persistent AF, and 39.1% (9/23) with long‐standing persistent AF. The mean left atrial diameter was 45.9 ± 9.5 mm (range 25–63).

TABLE 1.

Patient characteristics and procedural details.

Characteristic	Total (n = 23)
Age (years), mean ± SD	64.4 ± 7.99 (95% CI 60.9–67.9)
Male sex, n (%)	19 (82.6%; 95% CI 62.9–93.0)
Body mass index (kg/m2), mean ± SD	27.7 ± 3.99 (95% CI 26.0–29.4)
Type of AF, n (%)
Paroxysmal	8 (34.8%; 95% CI 18.8–55.1)
Persistent	6 (26.1%; 95% CI 12.5–46.5)
Long‐standing persistent	9 (39.1%; 95% CI 22.2–59.2)
Previous catheter ablation, n (%)	14 (60.9%; 95% CI 40.8–77.8)
Preoperative antiarrhythmic therapy, n (%)
Failed or intolerant to ≥ 1 Class I or III agent	16 (69.6%; 95% CI 49.1–84.4)
Amiodarone (any exposure)	9 (39.1%; 95% CI 22.2–59.2)
Flecainide	6 (26.1%; 95% CI 12.5–46.5)
Sotalol	2 (8.7%; 95% CI 2.4–26.8)
Dronedarone	2 (8.7%; 95% CI 2.4–26.8)
Documented intolerance to a Class I or III agent	3 (13.0%; 95% CI 4.5–32.1)
Beta‐blocker only, no Class I or III trial	7 (30.4%; 95% CI 15.6–50.9)
Declined antiarrhythmic therapy	1 (4.3%; 95% CI 0.8–21.0)
Neither Class I/III trial nor prior catheter ablation	5 (21.7%; 95% CI 9.7–41.9)
Left atrial diameter (mm), mean ± SD	45.9 ± 9.5 (range 25–63)
Left ventricular ejection fraction (%), mean ± SD	56.8 ± 7.9 (95% CI 53.4–60.2)
Coronary artery disease, n (%)	2 (8.7%; 95% CI 2.4–26.8)
COPD, n (%)	2 (8.7%; 95% CI 2.4–26.8)
Renal impairment, n (%)	2 (8.7%; 95% CI 2.4–26.8)
Mean procedural time (min), mean ± SD	92.9 ± 21.5 (95% CI 83.6–102.2)
Interrupted procedure, n (%)	1 (4.3%; 95% CI 0.8–21.0)
LAA exclusion, n (%)	21 (91.3%; 95% CI 73.2–97.6)
Complete box lesion set with confirmed entry and exit block, n (%)	22 (95.7%; 95% CI 79.0–99.2)
Electrical cardioversion required, n (%)	9 (39.1%; 95% CI 22.2–59.2)
Sinus rhythm at procedure end, n (%)	22/22 (100% of completed procedures)

Note: Completeness of the box lesion set was verified intraoperatively by demonstration of entry and exit block on epicardial three‐dimensional mapping (MapsPen, Medtronic), not by device tissue‐conductance feedback alone.

Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; LAA, left atrial appendage; SD, standard deviation.

More than half of the patients (60.9%; 14/23) had undergone previous catheter ablation. Preoperative antiarrhythmic exposure was heterogeneous (Table 1): 69.6% (16/23; 95% CI 49.1–84.4) had failed or were intolerant to at least one Class I or III agent, most commonly amiodarone (39.1%; 9/23) and flecainide (26.1%; 6/23), with documented intolerance to a Class I or III agent in 13.0% (3/23). The remaining 30.4% (7/23) had received a beta‐blocker only without a Class I or III trial, declined antiarrhythmic therapy, or had received amiodarone solely for cardioversion. Five patients (21.7%) had neither a Class I or III drug trial nor a prior catheter ablation. Mean left ventricular ejection fraction was 56.8% ± 7.9% (95% CI 53.4–60.2). Coronary artery disease, chronic obstructive pulmonary disease, and renal impairment were each present in 8.7% (2/23); prior ischemic cerebral events in 17.4% (4/23) and hemorrhagic events in 4.3% (1/23); at least mild mitral regurgitation in 43.5% (10/23) and tricuspid regurgitation in 21.7% (5/23).

The mean procedural duration was 92.9 ± 21.5 min (95% CI 83.6–102.2). One procedure was interrupted before completion because of anatomical limitations and tissue fragility, and the remaining 22 were completed. In all 22 completed procedures a complete posterior left atrial box lesion set was achieved and confirmed by demonstration of entry and exit block on epicardial three‐dimensional mapping (95.7% of the whole cohort; 95% CI 79.0–99.2). At the end of the procedure, all patients who underwent completed ablation were in sinus rhythm; electrical cardioversion was required in 39.1% (9/23) to restore sinus rhythm.

LAA exclusion was performed in 91.3% (21/23; 95% CI 73.2–97.6) of cases. It was omitted in two patients: in one the appendage was densely adherent and could not be safely dissected, and in the other the procedure was interrupted before the appendage was addressed because of frailty and tissue fragility. Both patients remained on long‐term oral anticoagulation.

3.2. In‐Hospital Outcomes and Complications

All 22 completed procedures were performed without intraoperative mortality, stroke, or myocardial infarction (Table 2). Conversion to sternotomy for bleeding control was required in three patients (13.0%; 95% CI 4.5–32.1), from the LAA in two and from the left upper pulmonary vein in one. One patient required transfusion of two units of red blood cells. All three patients recovered without further sequelae.

TABLE 2.

In‐hospital safety outcomes.

Outcome	n (%)
Pericardial effusion/tamponade	0
Conversion to sternotomy for bleeding	3 (13.0; 95% CI 4.5–32.1)
Blood transfusion	1 (4.3)
Pleural effusion requiring drainage	2 (8.7)
Pneumothorax	1 (4.3)
Pulmonary embolism	0
Pneumonia	2 (8.7)
Renal failure	0
Pacemaker implantation	0
Phrenic nerve palsy	0
Stroke/TIA	0
In‐hospital mortality	0
Hospital stay, mean (days)	6.5
Sinus rhythm at discharge, n (%)	20 (87.0; 95% CI 67.9–95.5)

Note: Three patients were in AF at discharge: the patient whose procedure was interrupted, who never achieved sinus rhythm, and two who reverted during the postoperative hospital stay.

Abbreviation: TIA, transient ischemic attack.

Pleural effusion requiring drainage occurred in two patients (8.7%), and pneumothorax in one (4.3%). Pneumonia developed in two patients (8.7%) and resolved with antibiotic therapy. No instances of pericardial effusion or tamponade, phrenic nerve palsy, renal failure, or pacemaker implantation were recorded. The mean length of hospital stay was 6.5 days.

At discharge, 20 patients (87.0%; 95% CI 67.9–95.5) were in sinus rhythm. Three patients were in AF: the patient whose procedure was interrupted, who never achieved sinus rhythm, and two who reverted to AF during the hospitalization following ablation. After an unsuccessful attempt at cardioversion, hemodynamically stable AF was tolerated until discharge in these two patients.

3.3. Follow‐Up and Long‐Term Outcomes

At the 1‐year follow‐up, 20 patients (87.0%; 95% CI 67.9–95.5) were in sinus rhythm, and all 20 were free of Class I or III AADs; two remained on beta‐blocker therapy, which was not counted as AAD use. The mean follow‐up duration was 23.2 months (Table 3).

TABLE 3.

Follow‐up outcomes.

Outcome	Value
Mean follow‐up duration (months)	23.2
Sinus rhythm at 1 year, n (%)	20 (87.0%; 95% CI 67.9–95.5)
Sinus rhythm without AAD at 1 year, n (%)	20 (87.0%; 95% CI 67.9–95.5)
Catheter touch‐up procedures, n (%)	3 (13.0; 95% CI 4.5–32.1)
Pace‐and‐ablate strategy, n (%)	3 (13.0; 95% CI 4.5–32.1)
Mortality at 1 year, n (%)	0 (0)
AFEQT quality of life (1‐year change; n = 21)
Symptoms domain, mean change (preop → 1 year)	+31.8 (59.9 → 91.7)
Daily Activities domain, mean change	+37.2 (49.1 → 86.3)
Treatment Concern domain, mean change	+13.1 (77.2 → 90.3)
Treatment Satisfaction domain, 1‐year score (no preoperative baseline)	81.7

Note: AAD denotes Class I or III antiarrhythmic therapy. Every patient in sinus rhythm at 1 year was free of Class I/III therapy; two remained on a beta‐blocker alone, which was not counted as antiarrhythmic drug use. All three domain changes exceed the published clinically important difference for the instrument (5 points overall; 7.1 for Symptoms, 5.1 for Activities of Daily Living). Scores were computed with the standard AFEQT algorithm, obtained during routine clinical follow‐up, and are reported as observed means without measures of dispersion.

Abbreviations: AAD, antiarrhythmic drug; AFEQT, Atrial Fibrillation Effect on QualiTy‐of‐Life.

Catheter touch‐up was required in three patients (13.0%) for recurrent atrial tachyarrhythmia at 9, 32, and 35 months; endocardial ablation restored rhythm control in all three. These patients had paroxysmal, persistent, and long‐standing persistent AF, one of each subtype; two had undergone previous catheter ablation, and all three had a complete block‐verified box lesion at surgery. Invasive electroanatomical mapping at reintervention showed that in two of them all four pulmonary veins and the posterior box remained isolated, the arrhythmia arising elsewhere—from antral and ostial left atrial signals in one, and in the other from a fibrotic posterior zone extending toward the mitral annulus that demonstrated rotational activity. In the third the pulmonary veins had reconnected and were re‐isolated ostially, with consolidation of the surgical box. Right‐sided substrate was addressed in all three: the superior vena cava was isolated in every case, with cavotricuspid isthmus ablation in one and right atrial micro‐reentry in another. A pace‐and‐ablate strategy was undertaken in three patients (13.0%) at 3, 6, and 13 months for refractory arrhythmia or bradycardia‐tachycardia syndrome; two had paroxysmal and one long‐standing persistent AF, one being the patient whose procedure was interrupted. Patient‐level detail for all six is given in Table 4.

TABLE 4.

Patient‐level detail of the six patients requiring rhythm‐related reintervention. ^a

Patient	Preoperative AF subtype	Prior catheter ablation	LA diameter (mm)	Complete block‐verified box lesion	Reintervention	Month after surgery
1	Paroxysmal	Yes	46	Yes	Catheter touch‐up	9
2	Persistent	No	47	Yes	Catheter touch‐up	32
3	Long‐standing persistent	Yes	38	Yes	Catheter touch‐up	35
4	Paroxysmal	No	55	No (procedure interrupted)	Pace‐and‐ablate	3
5	Long‐standing persistent	Yes	38	Yes	Pace‐and‐ablate	6
6	Paroxysmal	No	54	Yes	Pace‐and‐ablate	13

Electrophysiological findings at catheter touch‐up (CARTO electroanatomical mapping) b
Patient	Pulmonary veins	Posterior box	Substrate identified and ablated
1	All still isolated	Dense, intact	Rotational activity in a fibrotic zone extending from the posterior box to the mitral annulus, treated with a linear posterior lesion; superior vena cava isolation under phrenic nerve mapping
2	All reconnected	Consolidated	Ostial re‐isolation of all pulmonary veins; cavotricuspid isthmus ablation; superior vena cava ablation at the septal aspect, with immediate conversion to sinus rhythm
3	All four still isolated	Intact	Antral and ostial left atrial signals; right atrial micro‐reentry treated with linear ablation to the inferior vena cava; superior vena cava isolation, ablation toward the superior aspect abandoned owing to phrenic nerve capture

Abbreviations: AF, atrial fibrillation; LA, left atrial.

^a Patients are numbered for this table only and the order does not correspond to the chronological sequence of procedures. All three patients undergoing catheter touch‐up had a complete box lesion with demonstrated bidirectional block at the time of surgery.

^b Patient numbers correspond to the rows above. The superior vena cava was implicated in all three patients.

Outcomes stratified by preoperative AF subtype and by prior catheter ablation are shown in Table 5. These are presented for descriptive completeness only; as detailed in Section 2, the cohort has no meaningful power to compare subtypes, and no statistical comparison was performed.

TABLE 5.

One‐year outcomes stratified by AF subtype and prior catheter ablation. Descriptive only. As set out in the Methods, this cohort has no meaningful power to compare strata and no statistical comparison was performed. Percentages are given without confidence intervals because of the very small stratum sizes.

Outcome	Paroxysmal (n = 8)	Persistent (n = 6)	Long‐standing persistent (n = 9)	Prior catheter ablation (n = 14)	No prior ablation (n = 9)
Complete block‐verified box lesion, n (%)	7 (88)	6 (100)	9 (100)	14 (100)	8 (89)
Electrical cardioversion at procedure end, n (%)	1 (13)	3 (50)	5 (56)	6 (43)	3 (33)
Sinus rhythm at discharge, n (%)	5 (63)	6 (100)	9 (100)	13 (93)	7 (78)
Sinus rhythm at 1 year, n (%)	7 (88)	5 (83)	8 (89)	13 (93)	7 (78)
Sinus rhythm without AAD at 1 year, n (%)	7 (88)	5 (83)	8 (89)	13 (93)	7 (78)
Catheter touch‐up, n (%)	1 (13)	1 (17)	1 (11)	2 (14)	1 (11)
Pace‐and‐ablate, n (%)	2 (25)	0 (0)	1 (11)	1 (7)	2 (22)

Note: For context, and without implying comparison, a three‐center series of 475 patients reported freedom from atrial tachyarrhythmia of 72.7% in paroxysmal, 68.9% in persistent, and 54.2% in long‐standing persistent AF.

Abbreviation: AAD, antiarrhythmic drug.

AFEQT scores were available for 21 of 23 patients at 1 year; two did not return the follow‐up questionnaire, and rhythm status was available for both. Mean domain scores improved in all three domains assessed at both time points (Table 3), and each change exceeded the published clinically important difference for the instrument [8]. The Symptoms domain increased by a mean of 31.8 points (59.9–91.7), the Daily Activities domain by 37.2 points (49.1–86.3), and the Treatment Concern domain by 13.1 points (77.2–90.3). The Treatment Satisfaction domain, assessed at 1 year only, yielded a score of 81.7.

4. Discussion

This study describes outcomes of totally thoracoscopic AF ablation with the Gemini S clamp in 23 consecutive patients followed for a mean of 23.2 months. Sinus rhythm was restored in all completed procedures and maintained in 87.0% at 1 year. These findings are descriptive and are discussed below against published series and the limitations of this cohort.

4.1. Efficacy of Thoracoscopic Ablation

Sinus rhythm was present at the end of all completed procedures and in 87.0% of patients at 1 year, all of them free from Class I or III AADs, in a cohort in which 39.1% had long‐standing persistent AF and 60.9% had failed prior catheter ablation. Harlaar et al. reported closely comparable figures—84.7% in sinus rhythm at 1 year and 79.1% drug‐free—in a larger long‐standing persistent cohort [9].

Direct comparisons warrant caution. Our cohort contained many patients with long‐standing persistent AF and prior ablation failure, indicating advanced remodeling usually associated with poorer outcomes; the comparable success rate may reflect a standardized, electrophysiologically verified box lesion offsetting that disadvantage. Monitoring intensity also differs across studies and affects reported recurrence (see Section 5). The same caution applies to the time horizon: in the series cited above, sinus rhythm fell from 84.7% at 1 year to 62.9% at five, with Kaplan–Meier freedom from atrial arrhythmia of 50.0% at 5 years after a single procedure [9]. One‐year outcomes, ours included, say nothing about durability beyond that point.

Catheter touch‐up was required in 13.0% of patients, against 20.8% over 5 years in the largest long‐term series [9], though our follow‐up is shorter and the comparison of limited value.

4.2. Lesion Validation and the Mechanism of Late Recurrence

A distinguishing feature of this series is that lesion completeness was established electrophysiologically rather than inferred from device feedback, with entry and exit block demonstrated on epicardial three‐dimensional mapping in all 22 completed procedures. The alternative is unreliable: among 139 lesions predicted complete by tissue conductance, 5 (3.6%) were shown by formal block testing to harbor a gap, the right superior pulmonary vein being least reliably isolated [10]. We regard a physiological endpoint as a minimum standard for this operation rather than a refinement.

Because every patient who recurred underwent invasive mapping, we can say more than most series about what acute verification was worth. In two of the three, all four pulmonary veins and the posterior box were still isolated at reintervention, months to years after surgery: the lesion set had remained durable and the arrhythmia arose from substrate outside it. Only in the third had the veins reconnected. Recurrence here was therefore predominantly a problem of substrate beyond the surgical lesion rather than of the lesion failing—the same pattern reported with this clamp at 2‐year mapping [5] and in the largest long‐term series, where only 7 of 16 touch‐ups found a gap [9]. Where reconnection does occur, epicardial edema producing a false‐positive acute endpoint is one plausible explanation, and precisely why conductance‐based assessment fails [10].

Two practical conclusions follow. Epicardial fat should be dissected from the clamp footprint before energy delivery, and the clamp repositioned wherever apposition is incomplete. And recurrence should prompt early electrophysiological study rather than escalation of antiarrhythmic therapy, since in most of our patients the treatable substrate lay outside the surgical lesion and would not have been found otherwise. In the most advanced substrate, a staged hybrid strategy with planned endocardial validation may be reasonable from the outset; CONVERGE showed superior 12‐month freedom from atrial arrhythmia with a combined epicardial–endocardial approach over catheter ablation alone (67.7% versus 50.0%) in a population that, like ours, included many patients with long‐standing persistent AF [11].

4.3. Safety and Perioperative Outcomes

There was no perioperative mortality, stroke, or myocardial infarction. The 13.0% rate of conversion to sternotomy for bleeding, however, is substantially higher than in the larger published series and is the principal safety finding of this cohort. Conversion to sternotomy for bleeding occurred in 1.3% of 475 patients in a three‐center prospective series [12] and 1.6% of 558 patients in a single‐center safety analysis [13], and no conversion was required in the 77‐patient long‐standing persistent cohort of Harlaar et al. [9]. Our rate is therefore approximately eight to tenfold higher than the published experience.

With three events in 23 patients the estimate is imprecise (13.0%, 95% CI 4.5–32.1), and all three recovered without permanent sequelae. The conversions occurred in 2022 and 2023, distributed through the middle of a series running from January 2021 to August 2024 rather than confined to its earliest cases; a learning‐curve effect cannot be inferred from their timing, and we do not offer one. The bleeding sites—the appendage in two patients, the left upper pulmonary vein in one—indicate where this operation carries its principal hemorrhagic risk.

Pleural effusion (8.7%) and pneumonia (8.7%) resolved conservatively. Mean hospital stay was 6.5 days, longer than the median of 4 days in a 558‐patient series [13]. Against the 14.3% overall complication rate reported by Harlaar et al. [9], the total burden here was broadly similar though driven by a different event profile. There were no cases of phrenic nerve palsy, pacemaker implantation, or late pericardial complications.

4.4. Extent of the Lesion Set and the Role of Right Atrial Ablation

Our results, obtained with a left‐sided box and appendage exclusion alone, sit uneasily beside the biatrial lesion set of the Cox–Maze IV procedure, and whether right atrial lesions should be added deserves a direct answer.

The comparative evidence is conflicting. Adding right atrial lesions to a complete extended left atrial set did not improve 12‐month rhythm success in 305 patients (80.0% vs. 76.1%) but nearly doubled the pacemaker rate (16.5% vs. 7.5%) [14]. Conversely, in propensity‐matched concomitant open surgery a biatrial set gave better freedom from recurrence (91.9% vs. 85.3%), again with a much higher pacemaker rate (17.3% vs. 3.8%) [15], and with continuous event‐recorder monitoring biatrial ablation gave higher 12‐month freedom from AF (74.4% vs. 55.8%) [16]. The pacemaker signal is the most consistent finding: across 6135 valve surgery patients, implantation rose stepwise with the extent of ablation, from 3.0% after pulmonary vein isolation to 6.1% after left atrial and 12.8% after biatrial lesions [17]. No patient in our cohort required a pacemaker.

Two considerations specific to the stand‐alone setting weigh against routine right atrial ablation. The evidence above derives almost entirely from concomitant open procedures in which the right atrium is already exposed and the incremental cost of right‐sided lesions is small; thoracoscopically it requires additional dissection and instrumentation in patients whose indication is symptomatic AF rather than a structural lesion. More fundamentally, the cavotricuspid isthmus—the substrate of typical flutter and the most relevant right atrial target—cannot be reliably ablated epicardially and is addressed straightforwardly by an endocardial catheter. Retaining catheter access for the right atrium is therefore a rational division of labor rather than a concession, consistent with stand‐alone minimally invasive Cox‐Maze series in which extensive lesion sets deliver durable rhythm control at the price of greater complexity [18], and with the limited long‐term durability of catheter ablation alone in long‐standing persistent AF [19].

Our own reintervention data bear on this directly. Right‐sided substrate was addressed in all three patients undergoing catheter touch‐up: the superior vena cava was isolated in every case, one patient required cavotricuspid isthmus ablation, and one a right atrial micro‐reentrant tachycardia; in one, septal superior vena cava ablation converted the arrhythmia to sinus rhythm immediately. Three patients cannot establish a general claim, but the direction supports our position: right‐sided triggers and re‐entry do contribute to recurrence after a left‐sided box and are reached far more readily by catheter than thoracoscopically—superior vena cava isolation required phrenic nerve mapping in two patients, and in one, ablation toward the superior aspect was abandoned because of phrenic capture.

On the basis of these considerations, a complete, electrophysiologically validated left‐sided box lesion with appendage exclusion may represent a reasonable primary approach for stand‐alone thoracoscopic ablation across AF subtypes, with right atrial lesions reserved for documented right atrial macro‐reentry or typical flutter and delivered endocardially within a planned staged hybrid strategy. The present study has no comparison group and cannot establish that a left‐only lesion set is equivalent to a biatrial one; that requires a randomized comparison in the stand‐alone thoracoscopic population, which to our knowledge has not been performed.

4.5. Clinical Implications

The rhythm outcomes here are consistent with a role for thoracoscopic ablation between catheter ablation and open surgery for patients still symptomatic despite drug therapy or prior ablation, although the 13.0% conversion rate is not a modest procedural risk and must weigh in that positioning. With 60.9% having had prior endocardial ablation, the outcomes in that subgroup are consistent with an epicardial approach creating lesions difficult to reproduce endocardially, though with 14 patients this is an observation rather than a finding. Freedom from Class I or III AADs in every patient in sinus rhythm at 1 year is consistent with reduced dependence on antiarrhythmic medication; we did not measure drug‐related adverse events or healthcare utilization and make no claim about them. The AFEQT improvement, each domain exceeding the published clinically important difference, is consistent with rhythm benefit translating into patient‐reported benefit, in line with larger prospective studies [20, 21]. Uncontrolled before‐and‐after data in 21 patients cannot establish that the procedural risk is justified, but does provide device‐specific information useful in counseling patients.

The contribution of this series should be stated precisely, and it is narrow. It does not establish the validity of thoracoscopic ablation, demonstrated in cohorts of 475 and 558 patients and at 5 years [9, 12, 13] (Table S1); it does not establish that quality of life improves after the procedure, shown prospectively in 191 patients [20] and after hybrid ablation in 75 [21], and our findings are concordant with those rather than new; and with 23 patients, no comparator, and intermittent monitoring it establishes nothing about comparative efficacy. What it adds is device‐level evidence where almost none existed: the published experience with this clamp is a single three‐patient case series [5], and to our knowledge this is the largest reported cohort treated with it and the first to report 1‐year rhythm, safety, and patient‐reported outcomes together in a stand‐alone setting. Because bipolar clamps differ in energy delivery, jaw geometry, irrigation, and lesion architecture, device‐level data are not interchangeable with technique‐level data; the proper use of a series like this is to inform device selection and to be pooled in future meta‐analysis, not to guide choices between strategies. Whether this clamp's lesion architecture confers any advantage over other bipolar devices is a question the study raises and cannot answer.

5. Limitations

Several limitations must be acknowledged. The retrospective design is subject to selection bias and depends on existing clinical documentation rather than prospectively standardized data.

The sample of 23 patients limits precision: the 95% confidence interval around the primary outcome of 87.0% spans 67.9%–95.5%. For subgroup analysis the consequence is absolute rather than relative—as quantified in Section 2, power to detect even a large difference between paroxysmal and non‐paroxysmal AF is below 1%, so stratified outcomes are descriptive and must not be read comparatively. This cohort therefore cannot contribute to the comparison between thoracoscopic and catheter ablation within subtypes, which remains the central unresolved question in the field.

Most importantly for a rhythm study, monitoring was periodic 72‐h Holter ECG at 3, 6, and 12 months; intermittent monitoring systematically underestimates recurrence relative to continuous recording, and the success rate must be read accordingly. A 90‐day blanking period was applied per expert consensus [6], but events within it were not characterized. Antiarrhythmic withdrawal was individualized rather than protocol‐driven, introducing performance bias into the drug‐free endpoint.

Preoperative antiarrhythmic histories were heterogeneous and in part incompletely documented, and AF duration was often known only approximately; the cohort should not be characterized as uniformly drug‐refractory, and 21.7% of patients had neither a Class I nor III drug trial nor a prior catheter ablation. Quality‐of‐life data came from routine care rather than a protocol, were available for 21 of 23 patients, and are reported as observed means without dispersion; they are uncontrolled before‐and‐after observations. Anticoagulation after LAA clipping followed no standardized protocol beyond 3 months.

Finally, the 13.0% conversion rate substantially exceeds published series and its cause is not established; the timing of the three events does not support a learning‐curve explanation, and patient‐related anatomical factors cannot be excluded. As a single specialized center's experience, these findings may not generalize to lower‐volume centers or different institutional protocols.

6. Conclusions

In this consecutive single‐center cohort, totally thoracoscopic AF ablation with the Gemini S clamp achieved rhythm control across all AF subtypes represented, including in patients who had failed prior catheter‐based therapy. A complete box lesion with demonstrated bidirectional block was achieved in all completed procedures, and sinus rhythm was maintained in 87.0% of patients at 1 year, all of them free of Class I or III AADs. Three patients (13.0%) required conversion to sternotomy for bleeding, a rate exceeding larger published series that must be weighed against the rhythm outcomes. There were no strokes, phrenic nerve injuries, pacemaker implantations, or deaths. Given the sample size and absence of a comparison group, these findings are supportive rather than confirmatory.

These results are consistent with the established role of totally thoracoscopic ablation as an intermediate‐tier option between catheter ablation and open surgical treatment, and they extend that evidence to a device for which outcome data have been almost absent. They do not, on their own, establish comparative effectiveness. Future studies should include multicenter cohorts with extended follow‐up and continuous rhythm monitoring to validate long‐term rhythm durability. Direct comparative studies evaluating different surgical ablation technologies are needed to determine whether lesion transmurality and energy delivery influence outcomes in high‐risk populations, and a randomized comparison of left‐sided versus biatrial lesion sets in the stand‐alone thoracoscopic population would resolve the question of lesion‐set extent that the present data cannot address. Cost‐effectiveness analyses comparing thoracoscopic ablation with repeat catheter ablation or hybrid approaches would help refine patient selection.

Funding

The authors have nothing to report.

Disclosure

Registry and the Registration No. of the Study/Trial: N/A. This was a retrospective observational study and not a clinical trial; prospective registration was therefore not required.

Ethics Statement

The study protocol was approved by the local ethics committee (BASEC 2024‐02245) and the study was conducted in accordance with the Swiss Human Research Act (HRA), the Human Research Ordinance (HRO), and the Declaration of Helsinki.

Consent

All included patients had signed the institutional general consent permitting the further use of their health‐related data for research. No individually identifiable patient information is reported.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: The present cohort in the context of published thoracoscopic ablation experience.

Gruszczynski M., Rodriguez Cetina Biefer H., Rings L., Salzberg S. P., Ntinopoulos V., and Dzemali O., “Totally Thoracoscopic Atrial Fibrillation Ablation With the Gemini S Bipolar Clamp: A Single‐Center Descriptive Cohort Study of 1‐Year Rhythm, Safety, and Quality‐of‐Life Outcomes,” Journal of Arrhythmia 42, no. 5 (2026): e70498, 10.1002/joa3.70498.

Data Availability Statement

The data underlying this article are available from the corresponding author upon reasonable request.