Background Current evidence suggests that prolonged cardiac monitoring after cerebral ischaemia to detect paroxysmal atrial fibrillation (pAF) is likely to be cost effective, but the optimal duration of monitoring remains unclear.

Methods We performed a systematic review (studies published to December 2012) of all prospective and retrospective studies of rates of pAF early after TIA or ischaemic stroke in which consecutive patients underwent ≥12 hour cardiac monitoring, including unpublished data from our population–based study. Pooled estimates of rates of newly detected pAF were stratified by monitoring type and duration, study type, publication year and pre–defined pAF duration (any or ≥30 sec).

Results Among 19 retrospective and 21 prospective studies, the pooled rate of newly detected AF was 8.3% (851/10283; 95% CI 6.7–9.9, phet <0.0001). The rate tended to be higher in selected vs unselected populations (9.9%, 6.0–13.8 vs 7.9%, 6.2–9.5), but was unrelated to year of publication, and was similar in studies in which a duration of pAF ≥30 sec was required (8.6%, 6.3–11.0). However, duration of monitoring was the main determinant of the observed rate of pAF, accounting for 52% of all heterogeneity between studies in unselected populations. In stratified analyses, the rate of pAF among studies of unselected populations initially increased with duration of recording but plateaued at 5–7 days of monitoring (14.6%, 11.4–17.9, phet.=0.22), with no additional AF detected with 8–30 days of monitoring (13.7%, 9.8–17.5, phet.=0.17). Overall, 76.8% of patients with new pAF ≥30 sec were subsequently anticoagulated.

Conclusion Cardiac monitoring after TIA or ischaemic stroke detects clinically important rates of pAF in studies of unselected populations, with high rates of subsequent anticoagulation. A monitoring period of 5–7 days appears to be adequate.