Risk differences between atrial fibrillation patients with different clinical phenotypes: insights from the French population

Abstract

Abstract Background Atrial fibrillation (AF) patients represent a clinically complex, heterogeneous population comprising multiple homogeneous cohorts. Purpose We aimed to identify the common clinical phenotypes of AF patients and compare clinical outcomes between these subgroups. Methods A 1% representative sample of all AF patients hospitalized between 2010 and 2019 was identified from the French national database. Agglomerative hierarchical cluster analysis was performed using Ward’s method and squared Euclidian distance , to derive 4 distinct clusters of patients (Figure 1) using 31 clinical variables. Cox regression analyses were used to evaluate outcomes including all-cause death, cardiovascular death, non-cardiovascular death, ischaemic stroke, hospitalisation for heart failure (HF) and composite of ventricular tachycardia, ventricular fibrillation and cardiac arrest (VT/VF/CA). Results Four clusters were generated from the 12,688 patients included. Cluster 1 (n=2375) was younger, with a low risk of cardiovascular disease (CVD) and risk factors such as hypertension, dyslipidaemia and diabetes, but had the highest cancer prevalence. Clusters 2 (n=6441) and 3 (n=1639) depicted moderate-risk groups for CVD, with the female majority and high rates of lung conditions being the distinct features respectively. Cluster 3 also had the highest degree of frailty, with clusters 2 and 4 following closely. Cluster 4 (n=2233) represented a high-risk cohort for CVD with a high prevalence of coronary artery and vascular disease. Over a follow-up period of 2.0±2.3 years (median 1.1, interquartile range 0.1-3.4), cluster 3 had the highest cumulative incidence of all outcomes except for cardiovascular death which was highest in cluster 4. Compared to cluster 1, cluster 3 had the highest risk for all-cause death, HR 1.24 (1.09-1.41), cardiovascular death, HR 1.56 (1.19-2.06), non-cardiovascular death, HR 1.20 (1.04-1.38), hospitalisation for HF, HR 2.07 (1.71-2.50) and VT/VF/CA, HR 1.74 (1.20-2.53). Conclusion Four main clusters of AF patients were identified, discriminated by the differential presence of comorbidities. Our findings suggest that AF patients with moderate CVD-risk may have a poorer prognosis compared to AF patients with high CVD-risk in the presence of lung pathology. Therefore, this subgroup of patients may require more stringent management of existing comorbidities such as chronic obstructive pulmonary disease and sleep apnoea, alongside their AF.