Figure 3. Forest plot for the best-fitting conditional logistic
regression.
Discussion:
Widely used scores predicting CES risk in NVAF have emphasized clinical
characteristics including age and comorbidities. Few studies have
explored the contribution of LAA structural characteristics to
thrombotic risk. Clarifying the relationships between LAA anatomy,
physiology, and thrombotic risk may refine CES risk estimates and impact
anticoagulation or other stroke mitigation procedure decisions for NVAF
patients.
Previous examinations of LAA morphologic determinants of CES risk relied
on qualitative classifications of LAA shape, distinguishing between
appendages that resemble a Chicken-Wing, Windsock, Cactus, and
Cauliflower7. Several studies have demonstrated a
lower risk of stroke and transient ischemic attack, and a higher LAA
emptying flow velocities with Chicken-Wing type LAAs compared to other
LAA morphologic types7-9. The variable appearance of
LAA morphology in different TEE imaging planes, as well as LAA anatomic
complexity and heterogeneity, may limit the reproducibility and utility
of such morphologic criteria. Wu et al. examined inter-observer
agreements categorizing qualitative descriptions of LAA morphology by CT
in 2264 AF patients and found that all 3 reviewers came to consensus in
only 28.9% of scans10. Thus, LAA anatomic and
physiologic measurements that are objective and reproducible could be
valuable in predicting thrombus formation and thromboembolic risk in
NVAF.
Lower LAA emptying velocities are known to confer higher risk of
thrombogenesis8. The presence of AF during TEE and
persistent AF has predicted lower maximal LAA emptying flow velocities.
Petersen et al. observed higher flow velocities in patients with AF who
had Chicken-Wing LAA morphology, but interestingly only when patients
were in sinus rhythm at the time of TEE11. After
controlling for sinus rhythm at
the time of TEE, we observed lower LAA exit velocities in the LAA
thrombus cohort, consistent with prior publications.
The relationship between the LAA orifice area, a simple and reproducible
measurement on TEE and cardiac CT, and thromboembolic risk in NVAF is
unclear as previously published findings are conflicting. Khurram et al.
studied 678 patients undergoing AF ablation with pre-procedure cardiac
CT imaging, reporting an association of smaller LAA orifice size with
thromboembolic events in univariate analyses12. In
contrast, Lee JM et al. found that NVAF patients with stroke had larger
LAA orifice area and larger LA volume13. Similarly,
Lee Y et al. demonstrated larger LAA orifice diameters in patients with
CES14. These conflicting reports describing the
relationship of LAA orifice size and CES risk highlight the importance
of further investigation.
In contrast to prior work examining the relationship between LAA
structural characteristics and clinical thromboembolic event risk in AF,
our study examines the structure and function of the LAA in AF patients
with documented LAA thrombus. This difference in outcome of interest
compared to previous studies that used clinical history of stroke and
TIA as a surrogate for intracardiac thrombus formation represents a
unique aspect of our study insulating our results from the variability
associated with accurately diagnosing stroke or TIA. While most ischemic
strokes in patients with NVAF are cardioembolic, at least 24% may have
a non-cardioembolic etiology15. It may therefore be
advantageous to directly study patients with documented LAA thrombus
when characterizing LAA measurements that potentially predict thrombus
formation in NVAF. We observed a lower LAA peak exit velocity in the
thrombus group as compared to the control group (24.7 ± 1.6 cm/s vs.
37.9 ± 2.6 cm/s, p <0.0001). Additionally, we found that
patients in the thrombus group had smaller LAA OA at 0 and 90 degrees
(231.7 ± 13.0 mm2 vs. 328.6 ± 12.9
mm2, p<0.001), smaller LAA OA at 45 and 135
degrees (229.3 ± 12.6 mm2 vs. 336.4 ± 14.1
mm2, p<0.0001), smaller OA at the largest
diameter (239.0 ± 12.1 mm2 vs. 356.3 ± 13.5
mm2, p<0.0001), smaller aggregate OA (223.3
± 11.0 mm2 vs. 334.2 ± 12.4 mm2,
p<0.0001), and smaller maximum LAA depth (26.7 ± 0.7 mm vs.
29.4 ± 0.7 mm, p=0.0088) as compared to patients in the control group.
Our primary novel finding, smaller LAA ostial area as an independent
risk factor for LAA thrombus in NVAF, may help refine current CES risk
estimation models. In particular,
incorporating physiologic and anatomic measurements in clinical
thromboembolic risk models may be useful to identify truly low-risk AF
patients.
There were several limitations to this study. In this case-control
study, we identified patients with NVAF that were placed into the
thrombus and control groups from different populations of NVAF patients
undergoing TEE. This permitted identification of a larger number of
patients for the thrombus group but introduced selection bias into the
study. We undertook propensity matching in order to reduce the
possibility of confounding from differences in baseline characteristics
between the groups including age, LVEF, anticoagulation status at time
of TEE, and sinus rhythm at time of TEE. Despite the propensity
matching, a significant difference in the baseline characteristic of
LVEF persisted with the thrombus group having a lower LVEF compared to
the control group (45.1 ± 1.8% vs. 50.4 ± 1.5%, p=0.0285). However,
this association did not persist after best-fitting conditional
regression model analysis was performed (p=0.2161). Given the
retrospective identification of patients, LAA ostial diameter
measurements via TEE at all four mid-esophageal planes (0° and 90°, as
well as 45°and 135°) were not universally available. Aggregate OA was
calculated in order to incorporate maximal available measured LAA ostial
data. Aggregate OA was found to be significantly smaller in the thrombus
cohort as were all other OA evaluations including OA at 0° and 90°, OA
at 45°and 135° and OA at largest diameter adding validity to these
significant findings. Finally, the calculation of the OA used an ellipse
formula that does not factor in the irregular shape of the LAA ostium.
This limitation likely has less significance as both groups were
evaluated with the same formulaic assumptions.