Prolonged ECG monitoring after transient ischaemic attack for atrial fibrillation detection: feasibility of using meta-machine learning algorithm
J Tan, C Simela, T Habtezghi, S Maung, C Hayward, I Rehman, A Hassan, A Garcia, J Wu, C Gale, R NadarajahAbstract
Background
Whilst there have been several randomised clinical trials (RCTs) investigating prolonged ECG monitoring to detect atrial fibrillation (AF) after ischaemic stroke, this does not extend to transient ischaemic attack (TIA). (1–4) Early identification of AF post TIA and timely anticoagulation are crucial to prevent further stroke. (5,6) The FIND-AFDAS (Future Innovations in Novel Detection of Atrial Fibrillation After Stroke) framework may serve as a suitable enrichment criterion for a RCT evaluating prolonged ECG monitoring strategies after TIA. (7) The feasibility of a risk-guided AF detection RCT in patients with TIA still remains uncertain.
Purpose
To assess the use of prolonged ECG monitoring after TIA in routine practice and feasibility of implementing the FIND-AFDAS algorithm.
Methods
A retrospective analysis of patients (≥18 years) attending TIA clinics between April to August 2024 in a single tertiary regional stroke centre. TIA was defined by ‘focal unilateral motor deficit, speech/ language deficit or hemianopia, with symptoms duration <24 hours’. (2,8) All other presentations were excluded to reduce misclassification. Patients were subsequently stratified as high- and low-risk with the FIND-AFDAS algorithm using their routine health records. The primary outcome was the composite of new AF diagnosis, cerebrovascular event and all-cause death at one year. The secondary outcomes were the components of the primary outcomes and ECG monitoring strategy. Mantel-Haenszel tests were used for survival curve hazard ratio comparisons.
Results
Of 347 patients attending TIA clinics, 148 patients (42.7%) were diagnosed with TIA, of whom 17 had AF at baseline so were excluded. FIND-AFDAS score was calculated for the remaining 131 patients, with 26 (19.8%) identified as high-risk and 105 (80.2%) as low-risk. High-risk patients were older (median age: 80 vs 72; p<0.001) and had a higher prevalence of heart failure (19.2% vs 5.7%, p=0.041) compared to low-risk patients (Table 1). 120 (91.6%) received 72-hours Holter; with no patients receiving implantable loop recorder, wearable patch or external loop recorder. The primary composite outcome occurred in 4 (15.4%) high-risk patients and 6 (5.7%) low-risk patients (HR: 3.8; 95% CI 0.793 to 18.5) during 12-months of follow-up. AF diagnosis post TIA was infrequent, with 2 (7.7%) high-risk patients and 1 (1.0%) low-risk patients receiving an AF diagnosis by 12 months, with all subsequently prescribed oral anticoagulation. At 12 months, 23 high-risk patients were alive, had not had new AF detected and had not received cardiac monitoring longer than a 72-hours Holter.
Conclusion
AF diagnosis is rare after TIA event, and prolonged ECG monitoring not used. It is possible to identify high risk patients for AF, and there is still an unmet need to determine the appropriate prolonged ECG monitoring strategy in these high-risk individuals, and whether this can prevent further stroke events.Baseline characteristics and outcomesComposite outcomes