PRediction and detection of Occult Atrial fibrillation in patients after acute Cryptogenic stroke and Transient Ischemic Attack (PROACTIA)

Abstract

Background Studies with implantable cardiac rhythm monitors (ICRM) have shown that a 1/3 of patients with cryptogenic stroke/transitory ischemic attack (TIA) have episodes of subclinical atrial fibrillation (SCAF) as one of the potential risk factor for cerebral embolism. However, ICRM are costly and resource demanding. Purpose The PROACTIA study seeks to test a pre-specified scoring system to assess individual risk of SCAF in order to offer tailored therapy to patients after cryptogenic stroke/TIA. Methods Patients admitted with first time cryptogenic stroke or TIA were eligible for the study and underwent blood sampling, cerebral CT/MRI, carotid Doppler ultrasound, ECG, 24h-HolterECG, transthoracic transesophageal echocardiography, registration of medical history and implantation of ICRM during the index hospitalization The scoring system was composed of variables that have previously been found associated with AF occurrence: CHA2DS2-VASc, P-wave duration (P-dur), premature atrial contractions (PAC)/24h, supraventricular runs (SVR)/24h, left atrial end-systolic volume index (LAVI), and the biomarkers TnT, NT-proBNP and D-dimer. Data are presented as median (inter quartile range). Results Within 25 months, 434 patients screened and 251 patients were included in the present study and had an ICRM implanted. Eleven patients were later excluded yielding a study cohort of 176 patients with cerebral infarction and 61 with TIA that were followed for 833 (633–1028) days. AF was detected in 36%. It took 113 (25–336) days to detect AF, and 5 (2–14) days from AF-detection to initiation of NOAC. All variables were significantly increased in AF patients: no-AF vs AF: CHADS-VASC: 4 (3–5) vs 5 (4–6)*, LAVI mL/m2: 35 (28–40) vs 40 (35–50)**, PAC/24h: 69 (29–211) vs 347 (59–1917)**, SVR / 24-h: 1 (0–3) vs 3 (1–17)**, P-dur ms: 100 (100–120) vs 120 (100–120)**, D-dimer mg/L: 0,3 (0,2–0,6) vs 0,5 (0,3–0,9)**, TnT ng/L: 10 (7–16) vs 15 (9–27)**, NT-proBNP ng/L: 103 (49–283) vs 245 (102–774)**. *p<0.01, **p<0.001. Multivariate analysis yielded the following model: −8.524 + 0.057*LAVI (p<0.001) + 0.035*P-dur (p<0.001) + 0.873*LogPAC/24h (p<0.001). ROC analysis using leave-one-out cross validation: AUC=0.77. Applying the model to our population, it would identify a high-risk group (>80% true positive) consisting of 17 true positive and 4 false positive, and a low-risk group (<5% false negative) consisting of 17 true negative and 1 false negative. Conclusions ICRM detected SCAF in 36% of cryptogenic stroke/TIA patients within 27 months. LAVI, PAC/24h and P-duration were strong independent predictors of SCAF enabling a meaningful risk stratification that can be used for tailoring therapy in cryptogenic stroke/TIA patients. Funding Acknowledgement Type of funding source: Foundation. Main funding source(s): Stiftelsen Dam