We identified 407 patients with drug-refractory atrial fibrillation who underwent PVI between January 2008 and December 2014 at large military treatment facility. Baseline demographic data, procedural and lab details, and outcomes were obtained from manual chart abstraction of the electronic medical records (EMR). Patients with incomplete or unavailable imaging data were excluded from analysis (Fig. 1). Hypertension was defined as previous diagnosis, currently taking any anti-hypertensive medications, or a recorded systolic blood pressure > 140 mm Hg on 2 separate encounters. Congestive heart failure was defined as a previous diagnosis based on review of past medical history. Diabetes mellitus was defined as a previous diagnosis based on EMR review, currently taking any anti-hyperglycemic medications, or a most recent hemoglobin A1c value ≥ 6.5%. Vascular disease was defined as any previous history of myocardial infarction, coronary revascularization (percutaneous coronary intervention or coronary artery bypass grafting), or peripheral arterial disease. CVA was identified by EMR review and adjudicated utilizing the inpatient EMR or assessment by an outpatient neurologist evaluation. These risk factors, in addition to age and gender, were used to calculate a CHA₂DS₂Vasc score [11]. Post-PVI anticoagulation was recorded as the oral systemic agent on which the patient was discharged following PVI. Choice of oral systemic agent was at the discretion of the electrophysiologist performing the PVI. Post-PVI CVA events were identified utilizing the same criteria as outlined above. Patients with TIA by ICD-9 code listed following PVI were reviewed, but these instances were excluded unless deemed by the investigators to represent an embolic event. The vast majority of patients underwent pre-procedural transesophageal echocardiograph (89.5%) with finding of possible thrombus in one patient. CCTA in this patient performed with early and delayed phases was able to exclude thrombus. As the vast majority of pre-procedural CCTAs were obtained with only a single, early phase, all patients with LAA filling defects by CCTA were excluded from this analysis.

All PVIs were performed utilizing bilateral femoral venous access. For patients on warfarin therapy, this was continued without bridging anticoagulation in the peri-procedural period. In patients on NOACs leading up to PVI, these medications were held for 24 h prior to the procedure. Regardless of pre-procedural systemic anticoagulation agent, intravenous heparin was administered with a goal ACT of ≥350 s. ACT readings were reassessed at 30 min intervals and additional heparin boluses were administered as needed while catheters were in the left atrium. The esophagus was marked with a temperature probe and intracardiac echocardiography (ICE) was utilized for transseptal puncture. Transseptal puncture was performed using Brockenbrough needle under fluoroscopic and ICE guidance. Left atrial access was confirmed utilizing pressure transduction and saline administration. A multipolar mapping catheter was placed into the left atrium at operator discretion. Electroanatomic 3-D mapping of the left atrium was performed. Using contact force mapping, lesions were placed in a circumferential manner around the pulmonary veins. All ablation lesions were performed utilizing an open-irrigated catheter ablation system. The possibility of esophageal injury was minimized by performing ablation at < 30 watts along the posterior wall and terminating ablation with detected temperature rise in the esophageal probe. All other ablation lesions were performed utilizing 30–40 watts. Pacing at the right superior pulmonary vein was performed to confirm the absence of diaphragm stimulation prior to ablation at operator’s discretion. A multi-polar mapping catheter was then used to assess for any gaps. Immediate post-ablation isolation of the pulmonary veins was confirmed by LA and PV pacing to ensure bidirectional block. Non-PV triggers were induced utilizing isoproterenol and additional radiofrequency ablations administered as indicated. The ablation catheter was withdrawn to the right atrium. Screening for pericardial effusion was performed at the end of the case and pulmonary vein stenosis excluded using ICE. All catheters were then removed and, after reversal of anticoagulation with protamine, all sheaths were removed. All patients, regardless of CHA₂DS₂Vasc score, received 4 weeks of systemic anticoagulation (warfarin or a NOAC) following PVI. Choice of anticoagulant was at the discretion of the electrophysiologist performing the PVI.

From January of 2008 to March of 2011, images were obtained using a retrospective ECG-gated acquisition with a 64-slice CT scanner (Somatom Definition CT, Siemens, Erlangen, Germany) reconstructed at 0.7 mm slices centered on the 70% R-R interval. From March of 2011 to March of 2012, CCTA was performed utilizing a prospective ECG-triggered acquisition centered on the 60 ± 20% R-R interval. After March of 2012, a 128-slice dual-source scanner (Somatom Definition Flash CT, Siemens, Erlangen, Germany) utilizing prospective, ECG-triggered acquisition centered on the 60 ± 20% R-R interval was used. CCTA images were evaluated on a 3-dimension workstation (Vital Images, Inc., Minnetonka, MN). Left atrial (LA) and LAA parameters were obtained using electrophysiology planning software package. Manual segmentation of the LA and LAA was performed by delineating the borders utilizing axial reconstructions. LAA volumes were included in the LA volumes, however PV volumes were excluded. The LAA os was pre-determined to be the proximal border of the LAA and all visualized lobes were included in the volumes. LAA morphology (Fig. 2) was assigned by a single investigator (FK) based on review of volume rendered reconstructions and reported as chicken wing (CW), cauliflower (CF), cactus (CS), and windsock (WS) as previously described [10]. The CW was defined as a LAA morphology with a discrete bend in the proximal to mid-portion of the dominant lobe, folding back upon itself at some distance from the LAA ostium10. This morphology could also contain secondary lobes or twigs. The CF morphology was considered when LAA had a limited overall length and more complex internal characteristics. It was distinguished from other morphologies by variations in the ostium shape (oval vs round) and a variable number of lobes without an identifiable dominant lobe [10]. The CS morphology was classified as a dominant central lobe and secondary lobes that extend in both a superior and inferior direction [10]. Finally, the WS morphology was defined as a dominant lobe of sufficient length as the primary structure. Variations on this description to include different locations and numbers of smaller, secondary or even tertiary lobes were included in this morphologic group [10].