Multivariate analysis
In the binary logistic regression analysis, all variables with a univariate association with the presence of LA/LAA thrombus were entered into the model—namely, age, LA diameter, CHA2DS2VASc score, presence of diabetes, hypertension and vascular disease, as well as the stratification according to NTproBNP levels. Table 3 summarises the results of the multivariate analysis. It is clear from figure 1 that a NTproBNP level below the cut-off value was the most powerful predictor of the presence of LA/LAA thrombus (multivariate OR 25.016; p=0.016). The CHA2DS2VASc score and LA diameter were also independent predictors of LA/LAA thrombus (p=0.029 and p=0.023, respectively).