* Aggregate OA was calculated as the average of the measured OAs for the
subjects where all four mid-esophageal planes were available (0° and
90°, as well as 45°and 135°). In the case that only one pair of
orthogonal measurements (0° and 90° or 45°and 135°) then the available
value was used. 42 of the 61 subjects in the thrombus group and 48 of
the 61 subjects in the control group had all four mid-esophageal planes
available.
† TEE = transesophageal echo, BMI = body mass index, LVEF = left
ventricular ejection fraction, TIA = transient ischemic attack, LA =
left atrial, OA = ostial area, LAA = left atrial appendage
Table II summarizes candidate conditional logistical regression models
for the outcome presence of LAA thrombus. Model 1 included aggregate OA,
LAA exit velocity, and LVEF and was selected as the final model for
statistical results and inference based on the lowest AIC (27.578)
representing the best goodness-of-fit measurement among the candidate
models. Table III summarizes results from the best-fitting conditional
regression model (Model 1) examining the relationship of aggregate OA,
LAA exit velocity, and LVEF to the primary outcome presence of thrombus.
There was a significant association found between aggregate OA and
presence of LAA thrombus (p = 0.0038) as well as LAA exit velocity and
presence of LAA thrombus (p=0.0107) in the logistic regression model,
supporting these two LAA characteristics as risk factors for presence of
LAA thrombus. There was no association (p = 0.2161) between LVEF and
presence of LAA thrombus in the logistic regression analysis, suggesting
that LVEF does not represent an independent risk factor predictive of
LAA thrombus. These results are depicted graphically in Figure 3, a
Forest plot for the best-fitting conditional logistic regression.
Table II. Candidate Conditional Logistic Regression Models for the
Outcome (Thrombus or not)