Cardiac computed tomography following Watchman FLX implantation: device-related thrombus or device healing?

Abstract

Aims

Cardiac computed tomography (CT) is increasingly utilized during follow-up after left atrial appendage closure (LAAC). Hypoattenuated thickening (HAT) is a common finding and might represent either benign device healing or device-related thrombosis (DRT). The appearance and characteristics of HAT associated with the Watchman FLX have not been previously described. Therefore, we sought to investigate cardiac CT findings during follow-up after Watchman FLX implantation with a focus on HAT and DRT.

Methods and results

Retrospective single-centre, observational study including all patients with successful Watchman FLX implantation and follow-up cardiac CT between March 2019 and September 2021 (n = 244). Blinded analysis of CT images was performed describing the localization, extent, and morphology of HAT and correlated to imaging and histology findings in a canine model. Relevant clinical and preclinical ethical approvals were obtained.

Overall, HAT was present in 156 cases (64%) and could be classified as either subfabric hypoattenuation (n = 59), flat sessile HAT (n = 78), protruding sessile HAT (n = 16), or pedunculated HAT (n = 3). All cases of pedunculated HAT and five cases of protruding sessile HAT were considered as high-grade HAT (n = 7). Subfabric hypoattenuation and flat sessile HAT correlated with device healing and endothelialization in histological analysis of explanted devices.

Conclusion

Subfabric hypoattenuation and flat sessile HAT are frequent CT findings for Watchman FLX, likely representing benign device healing and endothelialization. Pedunculated HAT and protruding HAT are infrequent CT findings that might represent DRT.

Graphical Abstract

Open in new tabDownload slide

Introduction

Device-related thrombus (DRT) remains a significant complication following transcatheter left atrial appendage closure (LAAC) and appears on 1–5% of implanted devices, increasing the risk of thromboembolic complications.^1–5 DRT management involves an intensified course of anticoagulation and represents a clinical challenge due the significant bleeding risk of LAAC patients.¹ An accurate diagnosis and treatment of DRT is critical to avoid thromboembolic complications, whereas overdiagnosis of DRT might lead to irrelevant intensified anticoagulation with an increased risk of bleeding.

Usually, DRT is detected during routine device surveillance, typically 6–12 weeks after LAAC.¹

Although, transesophageal echocardiography (TEE) remains the most frequently used modality for follow-up imaging, cardiac computed tomography (CT) is increasingly utilized for both LAAC planning and follow-up.^6,7 An international expert consensus on the use of cardiac CT for preprocedural LAAC planning has been published.⁸ However, a similar consensus regarding the evaluation and surveillance of DRT on postprocedural imaging does not exist.

We previously described a CT-based algorithm to discern hypoattenuated thickening (HAT) on the atrial device surface as part of device healing or DRT after LAAC with the Amplatzer Amulet device (Abbott, Lake Bluff, IL).⁹

In the present study, we sought to investigate and characterize HAT as seen by cardiac CT on the Watchman FLX device (Boston Scientific Inc., Marlborough, MA). Patient CT findings were correlated with imaging and histological data from experimental canine Watchman FLX implantations.

Methods

Study design

This retrospective single-centre analysis included all patients with successful Watchman FLX implantation and 8-week follow-up cardiac CT at Aarhus University Hospital, Denmark, between March 2019 and September 2021. As per institutional protocol, patients were discharged with either dual antiplatelet therapy (DAPT) or aspirin (ASA) monotherapy. Patients with a glomerular filtration rate >30 ml/min were scheduled for follow-up cardiac CT.

Baseline and procedure data were obtained from a local prospective registry of all consecutive LAAC procedures performed at Aarhus University Hospital, Denmark (n = 885). Patient files were reviewed in cases with high-grade HAT to confirm treatment and subsequent follow-up imaging. Clinical outcomes were defined according to the Munich consensus document and retrieved from individual patient files and through a search in the regional patient registry.¹⁰ The study was conducted in accordance with the Declaration of Helsinki and approved by the Danish Data Protection Agency (1–16-02-419-16) and the Danish Patient Safety Authorities (3-3013-2736/1).

Pre-clinical animal data were provided by Boston Scientific, and examples of device healing and thrombosis were reviewed and compared to clinical findings. The preclinical protocol was approved by the Institutional Animal Care and Use Committee review board at American Preclinical Services (APS, Minneapolis, MN). The first, second, and last author had full access to all the included data and takes responsibility for its integrity and analysis.

Cardiac CT protocol and analysis

The CT acquisition protocol has previously been described.^8,9 In brief, prospective electrocardiogram-gated CT acquisition was executed using a high-pitch spiral protocol, iodine contrast, and a Siemens Somatom Definition Force scanner (Siemens Healthcare, Forchheim, Germany). A similar protocol was followed for preclinical scans. Additional details on CT acquisition can be found in the Supplementary material.

Images were analyzed using the integrated imaging software syngo.via (Siemens Healthcare, Forchheim, Germany) by three investigators blinded to clinical outcomes. The implanted device was identified and analyzed through multiplanar views (Figure 1).

Cardiac CT was evaluated for the presence of device-related HAT, visual peridevice leak, and distal LAA contrast patency (see Supplementary material).

HAT analysis

HAT was characterized based on localization, extent, and morphology. Localization was classified as either HAT on the device surface or isolated to the cove surrounding the screw hub. Extent was evaluated as any continuation onto the LA wall, the height of the HAT and of the screw hub cove. HAT height was measured from the atrial surface of the hypoattenuation to a line parallel to the bottom of the screw hub cove. The height of the cove was defined as the distance from this line to the level of the most proximal frame (Figures 1 and 2).

Subsequently, based on a visual assessment of shape and homogeneity, HAT morphology was characterized as sessile or pedunculated. Sessile HAT was further subdivided into flat sessile (≤3 mm) or protruding sessile (>3 mm) based on the height of the thickening (Figure 1). HAT in the screw hub cove was defined as isolated HAT with a height equal to or less than the height of the cove. Based on the design of the Watchman FLX device membrane, this was denominated as subfabric HAT (Figure 2).

The terminology and characterization of HAT is consistent with our previous study of HAT on the Amplatzer Amulet device but specifically modified according to the design differences between the Watchman FLX and Amulet.⁹ Pedunculated HAT and protruding sessile HAT without continuity onto the LA wall and/or with an irregular surface was considered high-grade HAT.

Preclinical data

The presented preclinical data (n = 5) is from a canine study (mongrel, 25–35 kg). Implantations were performed as per standard FLX protocol (See Supplementary material). After 45 days, the heart was explanted, and a gross necropsy was performed. The left atrium was opened and enface pictures of the atrial device surface were obtained. Subsequently, the heart was fixed in formalin and prepared for histopathology. The device and LAA tissue were plastic embedded before medial, lateral, and central sections were cut. Standard haematoxylin/eosin and trichrome staining was performed on all sections.

Statistical analysis

Data distribution was evaluated using QQ-plots and histograms. Continuous variables are presented as mean with standard deviation (SD) or median with interquartile range (IQR) and compared using the two-sample t-test, Wilcoxon rank-sum, or analysis of variance as appropriate. Categorical variables are expressed as counts (percentages) and compared using the Pearson’s χ² or Fisher’s exact test. Clinical follow-up data were analyzed using time-to-event analysis with Cox regression and presented as absolute events as well as incidence ratios. All statistical analyses were performed using STATA version 17 (StataCorp, College Station, TX).

Results

During the study period, a total of 297 patients underwent LAAC with the Watchman FLX. Eight-week cardiac CT follow-up was available in 244 individuals (82.2%), constituting the study cohort (Figure 3).

Mean age was 74.1 ± 8.8 years, 70.5% were males and the mean CHA₂DS₂-VASc and HAS-BLED scores were 4.0 ± 1.6 and 2.5 ± 1.1, respectively.^11,12 Additional baseline characteristics are presented in Table 1. A total of 53 patients did not have CT follow-up. In 16 (30.2%), this was due to renal failure, and in another 16 (30.2%), follow-up was not possible due to the COVID-19 pandemic (Figure 3). Median time from LAAC to follow-up CT was 55 days (IQR: 48–69), and the mean size of implanted devices was 27.5 ± 4.1 mm (Table 2). The majority were discharged with either DAPT (65.2%) or ASA monotherapy (32.8%) (Table 1).

HAT

HAT was seen in 156 (64%) of 8-week follow-up scans. In 59 (24.2%) of scanned patients, the HAT was classified as isolated subfabric hypoattenuation, constituting 38% of all HAT cases (Table 2). Among patients with HAT on the surface of the device (n = 97), 78 (80.4%) were classified as flat sessile HAT, 17 (17.5%) as protruding sessile HAT, and 2 (2.1%) as pedunculated HAT. In total, seven patients were classified as having high-grade HAT features. We found no significant association between baseline patient characteristics, implantation details, or discharge antithrombotics with any type of HAT (see Supplementary data online, Table S1).

Additional cardiac CT findings

In 19.7% of patients, a peridevice leak was identified on CT, displaying a median area at the mid-device level of 42.5 mm² (IQR: 26–97). Distal LAA contrast patency was present in 28 patients (11.5%) without any visible peridevice leak. Cardiac CT findings are summarized in Table 2.

Device-related thrombosis

Five patients with high-grade HAT (2 pedunculated, 3 protruding sessile) received intensified anticoagulation therapy for DRT based on their 8-week follow-up CT (see Supplementary data online, Table S2). The remaining two patients with high-grade HAT had protruding sessile HAT with either no LA wall continuity or an irregular surface (Figure 3, Table 3). Anticoagulation therapy was withheld in both due to multiple brain microbleeds and prior intracerebral haemorrhages.

Repeated follow-up imaging was scheduled by either cardiac CT (n = 4) or TEE (n = 1) in the patients treated with anticoagulation. Resolution was observed in three of five patients. One DRT patient died from pneumonia during extended follow-up, and another is awaiting further CT follow-up.

Of the 16 patients with protruding sessile HAT at 8-week follow-up, 10 patients (62.5%) were scheduled for additional follow-up imaging to monitor HAT development. No cases of flat sessile HAT were clinically considered as DRT (Table 3).

Mean follow-up time was 1.4 ± 0.8 years. The overall number of events during follow-up was low, and no statistically significant differences in adverse event rates were found between HAT categories (see Supplementary data online, Table S3).

Preclinical findings

Successful transcatheter Watchman FLX implantation was performed in all canines. On cardiac CT, a thin and smooth layer of HAT was observed across the surface of the device and in the screw hub cove, representing both subfabric and flat sessile HAT (Figure 4). One device (Figure 4, Row 5) presented with a contrast flow through the fabric.

Upon macroscopic evaluation, these findings correlated with a thin white pannus across the surface of the device, representing an ongoing healing process (Figure 4). Histology confirmed a thin layer of flattened cells with a typical endothelial morphology present across the device fabric for four of the five devices, while the remaining device (Figure 4, row 5) displayed only partial neoendothelization (Figure 4 and see Supplementary data online, Figures S1–S5).

Discussion

Cardiac CT has previously demonstrated increased sensitivity in detecting potential DRT when compared to TEE.⁹ Additionally, the increased spatial resolution of cardiac CT enables detection and evaluation of subclinical findings such as low-grade HAT. An accurate diagnosis of DRT is essential in LAAC, but discrimination between potential DRT or benign device healing can be difficult.

In this retrospective, single-centre analysis we found HAT in any form to be present across a majority (64%) of Watchman FLX implants at 8-week follow-up CT scan. We further propose a classification and clinical algorithm for evaluation of HAT supported by histological findings in experimental animals.

Subfabric hypoattenuation

In this study, subfabric HAT was observed in 24.2% of patients after Watchman FLX implantation. The Watchman FLX consists of a nitinol frame covered in part by a polyethylene terephthalate (PET) membrane. The individual nitinol struts are joined in a central screw hub, leaving a circular depression in the frame around the screw hub. Across the atrial device surface, the fabric is not tightly bound to each nitinol wire, resulting in a space underneath the fabric at the cove surrounding the screw hub. As the fabric is invisible on cardiac CT, contrast deficiencies in the screw hub cove can easily be mistaken for HAT across the atrial device surface. However, this cove is an intra-device space as clearly seen in our animal examples and subfabric hypoattenuation should not be mistaken for DRT (Figure 4). With increasing device compression, the fabric can even be lifted, further increasing the space between the fabric and the frame (Figure 2). Consequently, a straight line placed at the most proximal level of the device frame may underestimate the actual subfabric space. However, we propose this method as a conservative and reproducible approach.

Suggested algorithm

For this study, we developed an algorithm for the analysis and interpretation of HAT (Figure 5). While the categories presented in this algorithm seem useful in identifying potential DRT, patient specific factors such as hypercoagulability, bleeding risk and comorbidities should be considered before intensifying anticoagulation in patients with protruding sessile or even pedunculated HAT.² Furthermore, additional factors such as low-flow conditions, peridevice leak and implantation depth may be considered when deciding on increased anticoagulation or additional follow-up imaging.

Uneventful follow-up of patients with low-grade HAT on Amplatzer devices has previously been described, suggesting low-grade HAT to be a stage in the benign endothelization of the atrial device surface.⁹ While we lack paired human histology and imaging to confirm this, the canine data presented in this paper supports this interpretation.

Device healing

In the setting of LAAC, the relationship between device healing, HAT and thrombosis remains incompletely understood. Limited evidence exists on explanted human devices.^13,14 However, a series of early animal studies have been performed using both disc- and plug-type devices.^14–16 As displayed in these studies, neoendothelialization is a time-dependent process, gradually progressing during the initial post-procedural months. While Bass¹⁵ showed the surface of the disc-type Amplatzer Cardiac Plug (Abbott, Lake Bluff, IL) to be covered by stable neointima at 90 days, Kar et al.¹⁶ showed incomplete endothelization of the Amplatzer Cardiac Plug at 28 days post-procedure. In the latter study, the surfaces of all Watchman 2.5 devices were completely covered and incorporated at 28 days. Comparably, Schwartz et al. described complete coverage of the Watchman device at both 45 and 90 days.¹⁴ Although the healing pattern was similar to that in canines, complete endothelization was slower and less pronounced in humans.^13,14

These differences relating to device design as well as varying degrees of endothelization align with the pattern seen in the present analysis of HAT in a Watchman FLX population. At 8 weeks after LAAC, we found HAT on the surface of approximately 40% FLX implants, but only 6% HAT in a prior study of the Amulet device, suggesting that device healing might occur more rapidly for the FLX device.⁹

Device-related thrombosis

Experienced clinicians found five of the included scans to represent DRT and initiated intensified anticoagulation therapy. Two additional patients showed high-grade HAT changes, but a clinical history with multiple intracerebral haemorrhages and microbleeds lead to a decision of observation without further anticoagulation therapy. The observed rate of DRT in this study (5/244; 2%) align our results with rates found in large registries and trials.^{1,3,4,17–19}

Though this study was neither designed for nor powered to determine risk-factors for DRT, we found greater implant depths and peridevice leaks to be more common among patients with high-grade HAT (Table 3).

DRT remains a relatively rare complication and large registries are needed to substantiate these findings. However, the current evidence consistently underlines the continued importance of correct device positioning and thorough pre-procedural planning.^1,2,20,21

Lastly, no association was found between discharge antithrombotic therapy and DRT. This was despite a large proportion of patients receiving only single antiplatelet therapy (33.2%), supporting previous observations of sufficient protection using this approach.^18,22

Study limitations

Firstly, this is a single-centre study and thus represents the experiences within a high-volume LAAC setup at Aarhus University Hospital. Future studies will have to demonstrate the feasibility and translatability of the suggested algorithm.

As an observational study, elements of bias and confounding cannot be ruled out. However, analysis of all cardiac CT scans was performed blinded to the clinical interpretation and additional patient data to minimize this potential limitation. The low event rate during follow-up renders most statistical analyses of clinical outcomes underpowered. Also, most patients categorized as high-grade HAT were treated accordingly, potentially mitigating risk differences.

CT follow-up was only available for 82.2% of the study cohort, partially due to the reduced CT imaging capacity during the COVID-19 pandemic.

The study included only a limited number of canines with no representation of high-grade HAT.

Lastly, recent literature has shown a large proportion (∼20%) of DRT-cases to present later than six months post-implantation.¹ This study included only data on short-term follow-up (median 55.0 days), as further imaging is not routinely performed at our institution.

Conclusions

HAT is a frequent finding on CT imaging after LAAC with the Watchman FLX device. HAT can be classified into subfabric, flat sessile, protrouding sessile, or pedunculated.

Subfabric hypoattenuation and flat sessile HAT are frequent findings for Watchman FLX, likely representing benign device healing. Pedunculated HAT and protruding HAT are infrequent CT findings that might represent DRT and should warrant additional therapeutic considerations.

Prospective evaluation of the proposed algorithm and of the clinical impact of HAT is warranted.

Supplementary material

Supplementary materials are available at European Heart Journal - Cardiovascular Imaging online.

Acknowledgements

The authors wish to acknowledge the contributions of Athanasios Samaras (AS) in the early phases of analysis and his inputs to the initial configurations of the REDCap database.

Funding

Dr. Nielsen-Kudsk holds research grants from the Novo Nordisk Research Foundation (Hellerup, Denmark, NNF17OC0024868; NNF17OC0029510).

Data availability

The data underlying this article may be shared upon reasonable request to the corresponding author.

References

Author notes