Atrial Appendage Morphology
284 patients in the cohort had a cardiac CT or cardiac MRI performed for clinical indications. Three-dimensional structures of the left atrial appendge (LAA) were constructed using the volume-rendered post-processing technique. Standard measurements of LAA volume, os circumference, and diameters were obtained. The morphology of the LAA was also evaluated using multiplanar reconstruction. LAA morphologies were classified by the reading radiologist and confirmed by the primary investigator (SK). The LAA os was pre-determined to be the proximal border of the LAA and all visualized lobes were included in the volumes. LAA was reported as chicken wing (CW), cauliflower (CF), cactus (CS), and windsock (WS) as previously described13. Given the numbers and known stroke associations, we prespecified patients as CW (n=136) or non-CW (148) for analysis. In addition to stratifying by GRS, patients were stratified by the 4q25 variant of greatest significance to AF (r2200733 (NC_000004.12:g.110789013C>T) given the known association with left-right asymmetry. Patients were categorized at this SNV as carriers (C; n=109) if they were heterozygous (CT) or homozygous (TT). Patients with the CC genotype were classified as non-carriers (NC; n=175).

Statistical analysis

Continuous variables are expressed as mean ± SD and categorical variables. Analysis was performed using Prism (version 6.0d, GraphPad Software Inc, La Jolla, CA). Continuous variables were expressed as mean ± SD. The Student t  test and Mann‐Whitney U  test were used for the comparison of continuous variables between groups. The paired t  test and Wilcoxon signed‐rank test were used for the comparison of continuous variables between echocardiographic follow‐ups. The χ  2 test and Fisher exact test were used for noncontinuous variables. Bonferroni correction was used for multiple correction where applicable. A 2‐sided P  value <0.05 was considered statistically significant.