Atrial abnormalities predict silent atrial fibrillation in patients with crytogenic stroke with an internal loop recorder

Abstract Background Cryptogenic stroke (CS) causes about 30% of admissions to a stroke unit. Silent paroxysmal atrial fibrillation (PAF) is believed to be the underlying cause of a significant proportion of patients. The use of internal loop recorders (ILR) early after the CS has demonstrated benefits in the diagnostic yield, but the indication for ILR in the current guidelines remains unclear. Atrial contraction strain (ACS) evaluated by cardiac ultrasound could be of help to select the patients more prone to suffer from silent PAF. Methods Randomized prospective unicentric study to evaluate the usefulness of early ILR for rapid detection of silent PAF episodes in patients with CS. Clinical and ultrasound predictors of PAF occurrence (ACS) were studied in order to define patients needing a closer follow-up. The presence of left atrium (LA) subtel abnormalities (LA dilatation, maximum systolic global longitudinal strain <21%, atrial contraction strain <13%, atrial ejection fraction <55%) was used as a pre-specified variable before the randomization. Results Fifty-nine patients were included after suffering a CS (47.5% in the ILR group and 52.5% in de control group). There were no significant differences among groups regarding basal characteristics. The diagnosis of new silent PAF was made in 43.3% vs 7.1% of patients at 12 months follow up in the ILR group compared to the control group, respectively (0=0.002). In patients with a normal LA silent PAF was observed in 23% vs 7%, and in patients with an abnormal LA silent PAF was diagnosed in 59% vs 8% (ILR group compared to control group, respectively, p=0.007). Conclusions Atrial fibrillation episodes can be observed in more than 40% of patients when ILR is implanted early after the CS, increasing to almost 60% when ultrasound atrial abnormalities are detected. Therefore ILR should be implanted in all patients with an abnormal LA for early PAF episodes detection.