Background: The detection of paroxysmal atrial fibrillation (pAF) in TIA and stroke is resource intense with a range of monitoring devices available. Monitoring can be done during inpatient admission for stroke, but strategies optimised for use in TIA and minor stroke clinics are also necessary.
Methods: Patients recruited to the Oxford Vascular Study, who were in sinus rhythm at the time of presentation with TIA or minor stroke, underwent clinical assessment, neuroimaging, blood sampling, and 5-day ambulatory cardiac rhythm monitoring. Patients were split into derivation (n=300) and validation (n=300) cohorts. Using logistic regression we determined the predictive value for pAF of candidate variables identified from a systematic review of previous studies: age, NT-proBNP (<300/300-900/>900umol/L), left atrial dilatation, stroke severity and acute lesion on brain imaging.
Result: Among the 600 patients, pAF was identified in 46. Predictors in the validation cohort were age (reference <65; 65-75 – OR=1.12, 95%CI=0.31-4.11; age >75 - 2.95, 1.00-8.68), NT-proBNP (reference <300umol/L; 300-900umol/L - 2.19, 0.67-7.18; >900umol/L - 3.66, 1.03-12.99) and acute lesion on brain imaging (2.52, 1.03-6.16). A possible score based on these variables (age 65-75, 1 point, >75, 2 points; NT-proBNP >300, 1 point, >900, 2 points; acute lesion - 3 points) had reasonable predictive value in the validation cohort (C-statistic=0.79, 0.68-89, p<0.001; pAF rate=14% at score >2 and 2.2% at<2).
Conclusion: Using established risk factors it may be possible to improve prediction of pAF detection in TIA and minor stroke. With further validation, triaging patients for more urgent monitoring may be feasible.

PREDICTION OF OCCULT PAROXYSMAL ATRIAL FIBRILLATION IN PATIENTS WITH TIA AND NON-DISABLING STROKE