Distinct contributions of mu opioid and CB2 cannabinoid receptors to neuroimmune and behavioral responses.

Atrial fibrillation (AF) is one of the most common arrhythmias in adults and one of the most prevalent heart rhythm disorders (; ). Given its rising prevalence and increasing incidence with advancing age, AF has emerged as a global health epidemic, imposing substantial medical and economic burdens (; ; ). Recent guidelines have emphasized that the progressive nature of AF is a key consideration in its clinical management. The 2023 ACC/AHA/ACCP/HRS guidelines highlight the importance of early identification of patients at risk for AF progression for implementing appropriate therapeutic interventions and improving outcomes (). The progressive nature of AF underscores the importance of predicting disease progression to optimize treatment strategies. It is widely accepted that the natural history of AF follows a progressive course, initially characterized by non-sustained episodes triggered by ectopic activity (paroxysmal AF [PAF]) (). Over time, these episodes induce electrical remodeling within the atrial myocardium, which can subsequently promote or accelerate myocardial apoptosis and fibrosis (anatomical remodeling; ). Ultimately, alterations in the atrial myocardial substrate facilitate the persistence of AF through complex self-sustaining electrical activity, leading to persistent AF (PerAF). Previous studies have demonstrated that patients with PerAF at initial diagnosis exhibit higher mortality rates compared to those with PAF; patients with PerAF also demonstrate distinct electrophysiological characteristics (). Notably, PerAF shows stronger associations with severe adverse events, including stroke, systemic embolism, heart failure (HF) hospitalization, and other cardiovascular morbidity and mortality, compared to PAF (). Therefore, identifying the risk factors for PerAF has become increasingly important for optimal clinical management. Early detection of PerAF is crucial for optimal therapeutic intervention. While the 12-lead ECG remains the gold standard for arrhythmia diagnosis, its limited temporal coverage may fail to detect paroxysmal or asymptomatic AF episodes (). For patients with PAF, predicting their probability of progression to PerAF is necessary for adjusting subsequent treatment regimens. Hence, this study aimed to: (1) identify differential metabolites and relevant metabolic pathways between PAF and PerAF through targeted metabolomics; (2) investigate the associations between altered metabolites and clinical parameters; and (3) establish and validate an integrated machine learning (ML) model incorporating both metabolomic and clinical features for PerAF.