Author + information
- Received April 21, 2015
- Revision received July 6, 2015
- Accepted August 13, 2015
- Published online December 1, 2015.
- Adil Rajwani, MBChB, PhD∗∗ (, )
- Masoumeh G. Shirazi, MD∗,†,
- Patrick J.S. Disney, MBBS∗,
- Dennis T.L. Wong, BSc, MBBS, PhD‡,§,
- Karen S.L. Teo, MBBS, PhD∗,†,
- Sinny Delacroix, MD†,
- Ramesh G. Chokka, MD‡,
- Glenn D. Young, MBBS∗,† and
- Stephen G. Worthley, MBBS, PhD∗,†,‡
- ∗Department of Cardiology, Royal Adelaide Hospital, North Terrace, Adelaide, Australia
- †Division of Medicine, University of Adelaide, Adelaide, Australia
- ‡South Australian Health and Medical Research Institute, North Terrace, Adelaide, Australia
- §MonashHeart and Department of Medicine, Monash University, Melbourne, Australia
- ↵∗Reprint requests and correspondence:
Dr. Adil Rajwani, Department of Cardiology, Royal Adelaide Hospital, North Terrace, South Australia 5000, Australia.
Objectives Predictors of residual leak following percutaneous LAA closure were evaluated.
Background Left atrial appendage (LAA) closure aims to exclude this structure from the circulation, typically using a circular occluder. A noncircular orifice is frequently encountered however, and fibrous remodeling of the LAA in atrial fibrillation may restrict orifice deformation. Noncircularity may thus be implicated in the occurrence of residual leak despite an appropriately oversized device.
Methods Pre-procedural multislice computerized tomography was used to quantify LAA orifice eccentricity and irregularity. Univariate predictors of residual leak were identified with respect to the orifice, device, and relevant clinical variables, with the nature of any correlations then further evaluated.
Results Eccentricity and irregularity indexes of the orifice in 31 individuals were correlated with residual leak even where the device was appropriately oversized. An eccentricity index of 0.15 predicted a residual leak with 85% sensitivity and 59% specificity. An irregularity index of 0.05 predicted a significant residual leak ≥3 mm with 100% sensitivity and 86% specificity. Orifice size, device size, degree of device oversize, left atrial volume, and pulmonary artery pressure were not predictors of residual leak.
Conclusions Eccentricity and irregularity of the LAA orifice are implicated in residual leak after percutaneous closure even where there is appropriate device over-size. Irregularity index in particular is a novel predictor of residual leak, supporting a closer consideration of orifice morphology before closure.
The prevention of stroke is a key consideration in the management of atrial fibrillation. The left atrial appendage (LAA) is a major site of thrombus formation in atrial fibrillation, and as such is frequently implicated in cardioembolic events (1). Percutaneous closure of the LAA has an emerging role in those individuals for whom the attendant bleeding risks of oral anticoagulation (OAC) are prohibitive (2). The efficacy of LAA closure is based on an exclusion of this cul-de-sac structure from the circulation, and persistent communication may not adequately prevent the passage of emboli from the LAA. An accurate sizing of the LAA for the purpose of device size selection is paramount in this regard, but residual leaks occur in more than 40% of cases post-procedure despite meticulous multimodality imaging (3).
Current closure devices are circular in shape and are deployed at relatively low pressure. Computerized tomographic (CT) evaluation of the LAA has shown a highly variable shape of the orifice however, with significant eccentricity and irregularity shown in the majority of individuals (4). Moreover, a remodeling of the LAA has been reported in chronic atrial fibrillation that is likely to alter wall compliance and elasticity (5). A mismatch between orifice shape and circular occluder in the setting of a less pliable orifice may thus be implicated in the occurrence of residual leaks. Parallels can be drawn to procedures such as transcatheter aortic valvular implantation (TAVI) for example, where excessive eccentricity has been shown to predict perivalvular leak (6). We evaluated the associations of residual leak after LAA closure with eccentricity and irregularity of the orifice as assessed by pre-procedural CT.
All consecutive individuals with atrial fibrillation in whom an intra-appendage occluder device had been implanted at our institution were identified from a prospectively established registry. Eligibility for LAA closure required a confirmed indication for anticoagulation as determined by a CHA2DS2VASc risk score ≥2, and either a prohibitive risk associated with long-term OAC or the occurrence of embolic events despite therapeutic OAC. Pre-procedural contrast-enhanced CT imaging of the LAA was routinely acquired to exclude left atrial / LAA thrombus and to complement intraprocedural evaluation of LAA size and morphology by transesophageal echocardiography (TEE) and fluoroscopy.
LAA occlusion procedure
All procedures were conducted via femoral venous access under general anesthesia. Fluoroscopy, 2-dimensional (2D) TEE, and more recently 3D TEE were used for real-time guidance of trans-septal puncture and LAA occlusion. Closure device size was determined by consideration of orifice diameters derived from all imaging modalities including CT, TEE, and fluoroscopy. Deployment was in accordance with manufacturer guidance, and heparin was used to achieve an activated clotting time of >250 s. Comprehensive evaluation of the closure device by TEE was routinely conducted across all transducer-array planes immediately after final deployment to detect and thoroughly assess the magnitude of any residual leaks. All patients received aspirin 100 mg and clopidogrel 75 mg for 1 month post-procedure, which was then followed by aspirin monotherapy.
Pre-procedural evaluation of the left atrium, LAA, and neighboring structures was by multislice CT (dual source flash spiral 2 × 128 slice, Siemens Definition). Temporal resolution was 75 ms with a gantry rotation time of 0.28 s, detector collimation of 128 × 0.6 mm, pitch of 3.4, and tube voltage of 100 to 120 kV according to body habitus. Fifty milliliters of contrast (Omnipaque 350, GE Healthcare, Buckinghamshire, United Kingdom) was delivered via the antecubital vein with 50 ml saline flush at 5 ml/s. Automated contrast bolus tracking with a region of interest placed in the left atrium was triggered at 100 Hounsfield units. Delayed phase images were also acquired where LAA opacification was suboptimal to aid in the distinction of slow filling versus thrombus.
Retrospective evaluation of LAA orifice
Blinded assessment of the LAA orifice on CT was conducted retrospectively by a single investigator (A.R.). Multiplanar reconstruction was used to obtain orthogonal views of the LAA and to prescribe a double-oblique en face view of the orifice (Figure 1). The orifice was defined as the plane between the circumflex artery and a point 15 ± 5 mm from the tip of the limbus, reflecting as closely as possible the site where the proximal aspect of the closure device would be expected to lie. Thin-slab views were used for measurements, typically at 0.15- to 0.18-mm thickness. Measurements were made of the maximal diameter and orthogonal minor diameter, and freehand planimetry was used to obtain orifice circumference, area, and thus mean diameter. Occluder device circumference could be determined by [π × uncompressed device diameter], to determine whether an appropriate oversizing had been achieved in relation to orifice circumference. Left atrial volume on CT was calculated by the biplane area-length method [0.85 × 4-chamber area × 2-chamber area / atrial length] (7). Pulmonary artery pressure was estimated noninvasively from pre-procedural transthoracic echocardiogram. Blinded quantification of residual leak on procedural 2D TEE was conducted separately (M.G.S., P.D.). Color and spectral Doppler were used to identify the presence of residual peri-device leak, and leak width was measured at its waist.
Eccentricity and irregularity indexes
Eccentricity refers to the distortion of a circular orifice along a major axis. Eccentricity index of an elliptical orifice was quantified by the magnitude of difference between major and minor axes (Online Figure), and calculated as: [1 − (minor diameter/major diameter)]. A perfectly circular orifice thus has an eccentricity index of 0, but approaches 1 with increasing elliptical shape (Figure 1). Irregularity can be regarded as the global loss of symmetry of an orifice. Irregularity index was quantified as the departure in actual orifice area (areameas) from the predicted elliptical area (areapred = π × major semi-diameter × minor semi-diameter) (Online Figure), and was calculated as the absolute value of [1 − (areapred/areameas)]. A perfectly regular orifice thus has an irregularity index of 0, but approaches 1 with increasing irregularity (Figure 1).
Measurements and indexes are expressed as mean ± SD. Comparisons of clinical, atrial, LAA or device-related characteristics according to the presence or absence of residual leak was by t test or Fisher exact test as appropriate. Univariate logistic regression was used to identify predictors of residual leak as a binary outcome. Correlation between indexes of orifice morphology and the magnitude of residual leak as a continuous variable was by Spearman’s rank correlation coefficient and expressed as an rs value with 95% confidence intervals (CIs). Receiver operating characteristic (ROC) curves were generated to determine thresholds of eccentricity or irregularity at which residual leaks occurred. For this purpose two binary definitions of residual leak were tested, namely (1) the presence versus absence of any leak, and (2) significant leak ≥3 mm versus insignificant leak <3 mm. Area under the curve (AUC) was determined to assess discriminatory power of the index, and expressed with 95% CI and standard error of the mean (SEM). An optimal index threshold could then be selected that maximized sensitivity and specificity. Two-sided p values were used in all analyses, with a value <0.05 required to reject the null hypothesis.
Thirty-four consecutive patients underwent LAA closure with intra-appendage occluder devices, with a Watchman device (Boston Scientific Corporation, Marlborough, Massachusetts) deployed in 43% and a WaveCrest device (Coherex Medical, Salt Lake City, Utah) in 57%. In no case was occlusion unsuccessful or abandoned mid-procedure. Full procedural TEE images were available in all cases. Two patients had not undergone pre-procedural CT however, and LAA opacification proved markedly suboptimal in 1 patient, thus only the remaining 31 cases were evaluated. There were no procedural device-related complications in this cohort such as embolization, migration, or pericardial effusion. Nor were there any strokes, systemic embolic events, or cardiovascular deaths during a total 37.5 patient-years of follow-up.
Orifice and closure-device sizing
Mean circumference of the LAA orifice was 70.7 ± 12.2 mm, with a mean diameter of 22.5 ± 3.9 mm. Mean uncompressed device circumference was 83.6 ± 11.1 mm, consistent with a selection of device size that was on average 18.2% larger than the orifice. The uncompressed device circumference exceeded orifice circumference in all but 2 cases. No significant correlation was shown between device compression and either eccentricity or irregularity index.
In 18 cases (58%) there was no residual leak detectable by TEE. A leak with a width of <3 mm was present in 10 cases (32%), and a leak width of 3 to 4.9 mm was present in 3 cases (10%). None were found to have a severe leak ≥5 mm in width. A significant leak, defined as any leak ≥3 mm (3), thus occurred in 10% of all cases. Baseline characteristics according to the presence or absence of residual leak are listed in Table 1. On univariate analysis, predictors of a residual leak were eccentricity and irregularity indexes (Table 2). Device type, device diameter, and device oversize (analyses excluding those 2 cases where the device was undersized) were not predictive of residual leak; nor were orifice size, left atrial volume, pulmonary artery pressure, or comorbidities that might influence LAA remodeling such as hypertension and congestive cardiac failure (Table 1). Comparison of all imaging after collation of data confirmed that the predicted orifice closely matched the actual device site in all but one case.
Mean eccentricity index for the 31 cases was 0.17 ± 0.08 (range: 0.04 to 0.34). Eccentricity index was correlated with orifice size (rs = 0.42, 95% CI: 0.06 to 0.68, p = 0.02) (Figure 2). Eccentricity index was greater in cases with residual leak compared to those with no residual leak (0.21 vs. 0.14, p = 0.006) (Table 1). Increasing eccentricity was positively correlated with residual leak (rs = 0.48, 95% CI: 0.14 to 0.72, p = 0.006) (Figure 3). This association was not altered by the exclusion of those 2 cases where the device was potentially undersized. On ROC analysis, AUC for the presence versus absence of a residual leak was 0.75 (95% CI: 0.57 to 0.93, SEM 0.09) (Figure 4). A retrospectively derived eccentricity index threshold of 0.15 predicted the presence of residual leak with a sensitivity of 85% and specificity of 59%, thus with a likelihood ratio of 2.1. AUC for ROC curve of significant leak ≥3 mm versus insignificant leak <3 mm was similar at 0.76 (95% CI: 0.52 to 1.00, SEM 0.12) (Figure 4). An index threshold of 0.16 in this case predicted a significant residual leak ≥3 mm with a sensitivity of 100% and specificity of 50%, thus with a likelihood ratio of 2.0.
Mean irregularity index for the 31 cases was 0.03 ± 0.03 (range: 0.002 to 0.12). Unlike with eccentricity index, orifice size was not correlated with irregularity index (rs = −0.04, p = 0.80) (Figure 5). Irregularity index was greater in cases with residual leak compared to those with no residual leak (0.05 vs. 0.02, p = 0.0002) (Table 1). Increasing irregularity was positively correlated with the presence of residual leak (rs = 0.62, 95% CI: 0.34 to 0.80, p = 0.0002) (Figure 6). This association was not altered by the exclusion of those 2 cases where the device was potentially undersized. On ROC analysis, AUC for the presence versus absence of a residual leak was 0.86 (95% CI: 0.72 to 1.0, SEM 0.07) (Figure 7). A retrospectively derived irregularity index threshold of 0.04 predicted the presence of a leak with a sensitivity of 69% and specificity of 94%, thus with a likelihood ratio of 12.5. For prediction of a significant residual leak ≥3 mm versus insignificant leak <3 mm, AUC was 0.90 (95% CI: 0.80 to 1.00, SEM 0.06) (Figure 7). In this case an index of 0.05 predicted the presence of a significant residual leak ≥3 mm with a sensitivity of 100% and a specificity of 86%, thus with a likelihood ratio of 7.0.
In this retrospective analysis of LAA orifice morphology by CT in 31 individuals undergoing intra-appendage occlusion, eccentricity and irregularity of the orifice contour predicted the occurrence of residual leak and correlated with the severity of leak. Irregularity was a stronger correlate, with an irregularity index threshold of 0.05 demarcating a 7-fold increased likelihood of significant leak ≥3 mm in this cohort. These data thus support an interaction between LAA orifice morphology and the incidence of residual leak.
Device size and residual leak
The occurrence of residual leak has historically been thought to merely reflect an undersized device. Our protocol for LAA occlusion workup mandated a meticulous multimodality sizing of the LAA to avoid the risks of an incorrectly sized device. In the overwhelming majority of cases in this cohort, the circumference of the uncompressed occluder was appropriately oversized with respect to the orifice circumference. The mean oversize was 18.2%, commensurate with procedural practice where a device diameter is selected to exceed orifice diameter by at least 10% to 20% (8). Residual leak did not exceed 5 mm in any of the cases. The incidence of residual leak was also comparable to that found in other studies of intra-appendage closure (3). Finally, orifice and device sizes alone did not correlate with the occurrence of residual leak. These observations support an assertion that the occurrence of residual leak in contemporary practice is related to factors beyond a simple mismatch of occluder size to orifice size. The LAA might initially be assumed to deform fully to accommodate the contour of an oversized intra-appendage device. A remodeling of the LAA in chronic atrial fibrillation may result in a reduced capacity for deformation however, particularly in response to a compressible occluder that is deployed at relatively low radial force. LAA specimens from individuals with chronic atrial fibrillation exhibit significant dilation, stretching, and a reduction in pectinate muscle volume (9). In line with a well-recognized accrual of fibrosis in the left atrium in atrial fibrillation (9), histological analysis of the LAA showed endocardial fibroelastosis in most patients with chronic atrial fibrillation that was accompanied by significant endocardial thickening (5). The increase in collagen and elastic fiber disarray of the endocardial surface was observed to extend as far as the epicardium (10). These changes are likely to alter orifice elasticity and compliance of the LAA, compounding the challenge of occluding a noncircular orifice.
Orifice morphology in atrial fibrillation
Eccentricity and irregularity are frequently observed in analyses of the LAA. In a CT analysis of LAA morphologic parameters in 612 individuals a circular orifice was shown in only 5.7% of cases (4). The investigators observed an eccentric oval shape most frequently in 68.9%, and a more irregular shape was observed in the remaining 25.4%. In our dataset we also noted an association between orifice size and morphology, with increasing orifice size correlating with greater eccentricity. Chronic atrial fibrillation has been shown to be associated with an asymmetrical atrial stretch (11). This in turn would be expected to distort the LAA orifice with increasing eccentricity. Greater changes in LAA morphology have also been correlated with a more protracted burden of atrial fibrillation (12). Our identification of eccentricity and irregularity as predictors for procedural success will thus be relevant to a substantial proportion of patients submitted for LAA closure.
Eccentricity, irregularity, and residual leak
Eccentricity has been associated with a greater likelihood of incomplete apposition of a circular device to the walls of an orifice, for example, in TAVI (6). The strength of correlation between eccentricity and residual leak after LAA closure in our cohort was relatively modest however. This may reflect the more pliable orifice of the LAA in comparison to the calcific aortic annulus encountered in TAVI. Irregularity index is a novel measurement, however, that is particularly applicable to the highly variable morphology of the LAA. Moreover, irregularity was more closely associated with residual leak and may therefore be a particular consideration in procedural planning for LAA closure. We quantified irregularity by indexing the departure of the planimetered area of the orifice from the area that would be expected in a regular ellipse. Irregularity would arguably be better captured by a comparison of circumference rather than area. However, the predicted circumference of an ellipse is difficult to calculate, whereas the predicted area is readily determined. In the setting of significant eccentricity and/or irregularity, it might be reasonable to consider an ostial closure device rather than purely intra-appendage occlusion, for example, by the Amplatzer device (St. Jude Medical, Saint Paul, Minnesota). Alternatively, a more aggressive device oversize might appear logical. Interestingly, a greater degree of oversize was not statistically associated with a reduced occurrence of residual leak despite a numerically greater oversize in the group without residual leak. This particular measurement should perhaps be interpreted with some caution however, for example, where the final device position proves to be oblique to the true orifice. The role of a greater oversize target in the setting of an eccentric and/or irregular orifice may thus continue to merit further attention.
The retrospective nature of this study in a relatively small cohort can only allow the generation of a hypothesis; however, this is a logical process in the stepwise evaluation of an emerging technology. Eccentricity and irregularity indexes may simply be surrogate markers of an orifice that is more challenging to correctly size, although this does not necessarily diminish their potential utility. LAA morphology is complex, and the possibility must be acknowledged that other morphological features might have been even more influential in procedural success and predicting residual leak than orifice circularity. The results should not be extrapolated to non−intra-appendage devices, for instance, the Lariat device (SentreHEART, Redwood City, California). The proportion of cases with significant leak ≥3 mm was relatively small at 10%, which is likely to limit the strength of those ROC analyses where this particular threshold was used to dichotomize residual leak. In addition, the large number of univariate analyses may increase the possibility of a Type I error. The clinical significance of a residual leak is also incompletely understood. A retrospective review of the PROTECT-AF (WATCHMAN Left Atrial Appendage System for Embolic Protection in Patients with Atrial Fibrillation) study did not show a clear correlation between residual leak and the subsequent incidence of stroke (3). However, the investigators advocated caution in view of wide confidence intervals for hazard ratios, and a potential confounding by continued OAC in some patients with residual flow. Moreover, those protocols routinely prescribed OAC for at least 45 days post-procedure. Therefore, the prevention of residual leak after implantation of a device that is initially pro-thrombotic continues to seem important, particularly in contemporary patients in whom OAC typically cannot be safely administered in the first place. Thus, a residual leak of 3 ± 2 mm currently remains the consensus threshold used to define whether closure has been adequate (2).
Eccentricity and irregularity of the LAA orifice are frequently encountered during percutaneous closure, and are associated with an elevated risk of residual leak. Irregularity index in particular as quantified by pre-procedure CT is a novel predictor of peri-device leak. These parameters may in turn influence the choice of closure device type or may prompt a more aggressive device oversize. Larger prospective data are now appropriate to further validate the predictive value of these indexes, and potential influence on procedural success.
COMPETENCY IN MEDICAL KNOWLEDGE: Incomplete closure of the LAA is frequently noted after percutaneous implantation of occlude devices. Two indexes of orifice morphology (eccentricity and irregularity) are now identified that may assist pre-procedural prediction of residual leak, which may facilitate procedural skill and patient care during percutaneous LAA closure.
TRANSLATIONAL OUTLOOK: Predictors of residual leak warrant further validation in larger studies, in concert with further evaluation of their impact on clinical outcomes. In parallel, an evaluation is required of countermeasures in response to unfavorable orifice morphology such as the utility of ostial-sealing devices.
The authors acknowledge the technical support of Mr. Lars Kruse and Mr. Luke Moffat with respect to CT protocols and scanner specifications.
For a supplemental figure, please see the online version of this article.
Prof. Worthley has received fees for honoraria, consulting, and a proctorship from St Jude’s Medical; and received fees for honoraria and consulting from Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- area under the curve
- congestive cardiac failure
- confidence intervals
- computed tomography
- left atrial
- left atrial appendage
- oral anticoagulation
- receiver-operating characteristic
- standard deviation
- standard error of the mean
- transcatheter aortic valve implantation
- transesophageal echocardiography
- Received April 21, 2015.
- Revision received July 6, 2015.
- Accepted August 13, 2015.
- American College of Cardiology Foundation
- Stoddard M.F.,
- Dawkins P.R.,
- Prince C.R.,
- Ammash N.M.
- Meier B.,
- Blaauw Y.,
- Khattab A.A.,
- et al.
- Viles-Gonzalez J.F.,
- Kar S.,
- Douglas P.,
- et al.
- Cabrera J.A.,
- Saremi F.,
- Sanchez-Quintana D.
- Nucifora G.,
- Faletra F.F.,
- Regoli F.,
- et al.