AIS-defined insomnia symptoms and left atrial diameter index in patients with nonvalvular atrial fibrillation: a multicenter cross-sectional study

Insomnia symptoms may contribute to atrial remodeling, but their association with left atrial structure in established nonvalvular atrial fibrillation remains unclear. We examined whether Athens Insomnia Scale-defined sleep status was associated with body surface area-indexed left atrial diameter (LADi). This multicenter cross-sectional study included 748 patients with nonvalvular atrial fibrillation from three centers. Patients were classified as having no sleep disturbance, suspected insomnia, or insomnia. Continuous LADi was the primary outcome, and elevated LADi (≥ 2.4 cm/m²) was the secondary outcome. Multivariable linear and logistic regression analyses were supplemented by sensitivity analyses, subgroup analyses, and propensity score overlap weighting. The groups included 284, 306, and 158 patients, respectively. Mean LADi increased across the three groups (2.22 ± 0.46, 2.36 ± 0.55, and 2.49 ± 0.58 cm/m²; P < 0.001), as did the proportion with elevated LADi (26.8%, 38.6%, and 48.1%; P < 0.001). In the fully adjusted model, insomnia was associated with higher continuous LADi (β = 0.114, 95% confidence interval: 0.017–0.211; P = 0.021), whereas the association with elevated LADi was not statistically significant (odds ratio = 1.547, 95% confidence interval: 0.952–2.514; P = 0.078). Estimates were generally directionally consistent in sensitivity analyses but were attenuated and no longer statistically significant after excluding patients with heart failure and after propensity score overlap weighting. Athens Insomnia Scale-defined insomnia was associated with modestly higher continuous LADi, but not with elevated LADi as a dichotomous outcome. The attenuation after excluding patients with heart failure and after overlap weighting suggests that the association was not consistently robust and may partly reflect cardiovascular comorbidity. The cross-sectional design precludes causal inference.