Introduction: Traditional predictors of paroxysmal AF (PAF) including left atria size, CHADSVASC score or rate of atria premature beats are inaccurate to predict PAF in elderly stroke patients. We aimed to evaluate the yield of atrial dysfunction markers such as NT-ProBNP and Left atrial deformability in acute elderly cryptogenic stroke patients to predict PAF.
Methods: We prospectively evaluated cryptogenic stroke patients <72 hours from stroke symptoms. All patients underwent NT-ProBNP level measurement and left atrial strain (LAS) assessment by speckle tracking echocardiography. We performed a continuous wearable Holter monitoring from the admission up to 4 weeks after stroke.
Results: Out of 150 cryptogenic stroke patients, one hundred and ten patients fulfilled 4 weeks of Holter monitoring. Median age was 75 years (69-83).The percentage of PAF detection was 24.5%. Age, sex, hypertension, diabetes, CHADSVASC score, NIHSSS, left atria volume, left ventricle ejection fraction and rate of atria premature beats were comparable between patients with and without PAF detection. Admission NT-ProBNP levels were higher (346.9 pg/ml [175-685]) in whom PAF was detected compared to those who did not (203.5 pg/ml [76.75-453.95] (p=0.028). Conversely, LAS was lower 18.66% [13.98-26.71] in PAF compared to no PAF detected patients 27.43% [21.28-36-56] (p=0.003). Both markers were inversely correlated (Spearman Rho -0.31) (p=0.007). The AUC of NT-ProBNP was 0.65 (0-53-0.76) with cut-off point of NT-ProBNP> 102 pg/ml (sensitivity 96%, specificity 33%) and LAS was 0.73 (0.59-0.86) with cut-off point of LAS <19.28% (sensitivity 50%, specificity 80.4%). The combination of both atrial dysfunction markers showed an AUC of 0.76 (0.69-0.88) with cut-off point of 0.23 (sensitivity 94%, specificity 50%).
Conclusions: The combination of both NT-ProBNP and LAS markedly improve the predictive accuracy for PAF detection in elderly cryptogenic stroke patients.