Atrial myopathy, characterized by electromechanical atrial remodeling, can lead to development of atrial fibrillation (AF) and is independently associated with embolic stroke [1]. Approximately 20%-40% of all strokes are classified as cryptogenic, which are associated with a 3%-6% risk of recurrent stroke within 1 year [2-4]. AF has been detected in 12%-38% of patients within 1 year of cryptogenic stroke supporting the possibility of a prominent cardioembolic mechanism [2,5]. Oral anticoagulant medications, unfortunately, have not demonstrated superiority to antiplatelet agents in preventing recurrent stroke in all patients with cryptogenic stroke or in enriched populations of patients with cryptogenic stroke and suspected atrial myopathy [3,4,6,7].

There is no consensus on the optimal markers to detect atrial myopathy clinically for purposes of stroke prediction. We aimed to determine if left atrial (LA) mechanical dysfunction, measured by left atrial emptying fraction (LAEF), is associated with AF and recurrent stroke or transient ischemic attack (TIA) in patients with cryptogenic stroke.

This study was approved by the Pennsylvania State University Institutional Review Board (IRB no. 00017160) and follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [8]. Patients with cryptogenic stroke (March 2015-March 2022) who received implantable loop recorders (ILRs) for AF screening with at least 1 year of follow-up and interpretable echocardiograms were included (n=192). Patients were excluded for missing covariate data (n=34) resulting in a final cohort of 158 patients. All 158 patients had ILRs present for at least 1 year or developed AF within 1 year. LAEF (exposure variable) was calculated from echocardiograms by measuring the maximal change in atrial volume over the course of the cardiac cycle. AF (primary outcome) was defined by at least 30 seconds of AF on ILR. Data on recurrent stroke or TIA (secondary outcome), covariates, and medication use were obtained by review of the electronic medical record. The covariates included in this study included age, sex, hypertension, coronary artery disease (CAD), prior stroke or TIA, peripheral artery disease, diabetes mellitus, congestive heart failure, carotid artery stenosis, and left atrial volume index (LAVI).

The relationship between LAEF and AF was evaluated using a restricted cubic spline, cumulative incidence curves, and multivariable Cox proportional hazards models. Model discrimination and calibration were evaluated by calculating the C-statistic and Hosmer-Lemeshow χ² statistic. The association between LAEF and recurrent stroke or TIA after cryptogenic stroke was evaluated using Cox proportional hazards models.

Additional details on covariate definitions and statistical analysis are available in Supplementary Methods.

Of the 158 patients (mean age, 70.2 years; 52.5% female) in our cohort, 43 had AF (27.2%) over a mean (standard error [SE]) follow-up time of 282 (10.8) days. Compared to those who did not, patients who developed AF had greater LAVI (36.76 mL/m² vs. 31.01 mL/m², P<0.01), lower LAEF (48.21% vs. 56.93%, P<0.01), and were less likely to have diabetes (P=0.02). There were no additional significant differences in baseline parameters between groups (Supplementary Table 1).

A restricted cubic spline demonstrated a significant linear association (P for non-linearity=0.82, P for association=0.05) between decreasing LAEF and an increased incidence of AF after cryptogenic stroke (Figure 1A). The cumulative incidence of AF was lowest in patients with LAEF ≥53% (Figure 1B). LAEF <40% was independently associated with a 2.81-fold (1.09-7.23) increased risk of AF after cryptogenic stroke after adjustment for all covariates. When modeling as a continuous variable per 5% decrease, this association was no longer statistically significant after adjusting for LAVI. The addition of LAEF (Model 3) to a model constructed with all covariates (Model 3’) resulted in improved discrimination with acceptable calibration (Table 1).

Recurrent stroke or TIA was identified in 11 of the 158 patients (6.9%) with a mean (SE) time from initial stroke to recurrent stroke of 171 (37.2) days. Compared to those who did not, patients who developed recurrent stroke or TIA had lower LAEF (44.79% vs. 55.28%, P=0.02) and were more likely to have CAD and develop AF (P=0.03) after the index stroke. Those who were on therapeutic anticoagulation at the time of stroke had reduced stroke severity as determined by the lower National Institutes of Health Stroke Scale. There were no additional significant differences in baseline parameters (Supplementary Table 2). LAEF per 5% decrease was independently associated with an increase in recurrent stroke or TIA. This association was attenuated after adjustment for LAVI and AF (Table 1).

The present study demonstrates that in patients with cryptogenic stroke, reduced LAEF is an independent risk factor for AF and may improve prediction of AF above what can be achieved using established stroke risk factors including LAVI. Our results indicate that the association between reduced LAEF and recurrent stroke or TIA may be mediated by LA size and development of AF.

Oral anticoagulation for the prevention of recurrent stroke in patients with cryptogenic stroke is currently guided by detection of AF post-stroke [9]. Detection of atrial myopathy, an upstream phenomenon independently associated with AF development and embolic stroke, may help refine this strategy. There is no consensus on the best variables for detection of atrial myopathy clinically, although several recent studies have indicated that analysis of LA mechanics is superior to analysis of LA size alone [10]. The four randomized trials evaluating the efficacy of oral anticoagulation for prevention of recurrent stroke in patients with cryptogenic stroke did not incorporate functional LA parameters or LA volume measurements to select candidates for anticoagulation [3,4,6,7]. If our findings are validated, future studies should consider analysis of LA mechanics to guide anticoagulation for prevention or reduction in severity of recurrent stroke after cryptogenic stroke.

There are limitations to be considered in this study. Only 11 of 158 (6.9%) patients suffered recurrent stroke or TIA. Thus, our analysis was largely exploratory for this endpoint. Additionally, despite statistical adjustments being made for potential confounding variables, imperfectly measured and unmeasured variables were not accounted for, as is the case in observational studies.