Impact of left atrial appendage morphology on thrombus formation in TAVI patients with atrial fibrillation

This single-centre cross-sectional study retrospectively included patients with symptomatic aortic valve stenosis (AS) that underwent a pre-procedural CT scan before trans-catheter aortic valve implantation (TAVI). We screened 503 patients between January 2016 and January 2018 at a local university hospital. Entry criteria included existing pre-procedural CT scan between 1st of January 2016 and 1st of January 2018 and diagnosed atrial fibrillation. The resulting collective consisted of 231 patients (flow chart in Fig. 1). The patients’ medical histories were retrospectively collected using their chart reviews in the clinical information system (SAP, Walldorf, Germany). Age, gender, BMI, history of hypertension, diabetes, heart failure, cardiomegaly, aortic valve replacement (AVR), and the presence of a cardiac pacemaker were recorded. CHA2DS2-VASc scores were calculated with one point assigned for a history of congestive heart failure, hypertension, and 74 years ≥ age ≥ 65 years, female, diabetes mellitus and vascular disease and two points assigned for age ≥ 75 years, a history of stroke or transient ischemic attack (TIA), with a maximum score of nine. Demographic and clinical data are given in Table 1. The study was approved by our institutional review board. Patients gave written informed consent for the clinical examinations.

All patients underwent a retrospectively ECG-gated spiral MDCT of the aortic valve region (Somatom Definition Flash, Siemens Medical, Erlangen, Germany). The CT scans were performed with a tube voltage of 100 kV and a tube current of 300 mAs. We used Ultravist 300 (Bayer, Berlin, Germany) as intravenous contrast agent. The injection protocol included a test bolus of 15 ml, followed by an injection of 95 ml with an injection rate of 6.0 ml/s. Images were reconstructed by an iterative reconstruction kernel (I39f) with a slice thickness of 0.75 mm with automated detection of best-systolic and best-diastolic time-points of the R–R interval. Additional reconstructions were performed at every ten percent of the R–R interval.

CT images were stored in our picture archiving system (Centricity, GE-Healthcare, Milwaukee, WI, USA). Image analysis of arterial and venous phase images was performed in consensus by two experienced examiners. LAA thrombus was defined as a persistent and significant filling defect in both contrast phases (Fig. 2). The left atrial diameter was assessed in the axial plane vertically to the aortic valve. We categorized the LAA anatomy into four different morphologies (Fig. 3): chicken-wing, windsock, cauliflower, and cactus, based on Wang’s classification [5].

We documented neuro-embolic events depending on the LAA thrombus’s presence within a follow-up of 18 months after TAVI. Post-procedural neuro-embolic events were defined as a hospital re-admission due to an acute, ischemic, neurological event within 18 months after the pre-procedural CT scan.

Within the seven patients with chicken-wing configuration and LAA thrombus, 42.9% (n = 3) experienced a neuro-embolic event within 18 months after TAVI. In contrast, 20.9% (n = 9) of the 43 patients with non-chicken-wing configuration and LAA thrombus, experienced a neuro-embolic event within 18 months after TAVI (OR: 0.35, 95%CI 0.67 to 1.87, p = 0.34).

The LAA derives from the primordial left atrium with a reservoir function in conditions of fluid overload [20]. In patients with non-valvular atrial fibrillation, it is with over 90% the major location for intra-cardiac thrombus [2, 20]. Previous studies suggest that the LAA morphology correlates with the risk of thrombus and stroke [3, 7‐9, 13, 16]. However, no statistical discrimination between both risks is available so far. The present study suggests that the LAA morphology might have an impact on the development from a LAA thrombus into an embolus with resulting neuro-embolic event. In our study, patients with chicken-wing morphology had a lower risk to develop a LAA thrombus. However, once a thrombus is formed patients with chicken-wing LAA are more likely to develop a neuro-embolic event.

Published LAA-classification schemes show inter-observer and intra-observer variabilities with limited reproducibility [7, 9, 15]. Our data suggest that it is crucial to differentiate chicken-wing and windsock morphologies. Since windsock showed the highest thrombus-rates and chicken-wing, the highest neuro-embolic event rate in patients with LAA thrombus. Whereas, from our experience, both chicken-wing and windsock configuration are easiest to detect. Correct identification of these two configurations could be a useful tool for radiologists.

The main limitation was the monocentric study design. Moreover, our highly specific patients population with AS and atrial fibrillation, results in relatively small number of patients. In addition, common LAA-classification schemes show inter-observer and intra-observer variabilities with limited reproducibility.