Abstract
Aortic aneurysms (AAs) are influenced by diverse modifiable and non-modifiable risk factors. The underlying heterogeneity in this patient population complicates efforts to optimize risk stratification. Phenomapping leverages unsupervised clustering algorithms to group similar patients based on phenotypic profiles and can identify homogenous clusters in complex data patterns. While clinical classification typically relies on aneurysm morphology and presentation, unsupervised phenomapping could uncover alternative, data-driven risk profiles. Unsupervised clustering (k-means) utilizing data from the UK Biobank was conducted. All patients recorded with aortic pathologies (ICD-10 group I71) were included. Clustering variables encompassed demographic and clinical parameters, emphasizing relevant comorbidities. The main analyses were conducted in abdominal AA, whereas validation analyses were conducted in the remaining aortic pathologies. The primary outcome was all-cause mortality. The study population consisted of 4623 participants (21.2% female), including abdominal (66.9%), thoracic (25.4%), and thoracoabdominal AA (0.5%), as well as aortic dissections (7.2%). Cardiovascular comorbidity emerged as the primary driver of cluster differentiation both in the main and validation analysis outweighing morphological differences. Survival analysis highlighted higher all-cause mortality rates in these multimorbid clusters. Cardiovascular and renal comorbidities can be leveraged in an informative comorbidity index. The non-linear increase in mortality risk may be indicative of the superimposed effects of the aortic pathology itself. Cardiorenal multimorbidity appears to be the most relevant separator between AA patients. However, the burden of disease associated with an aortic pathology itself relevantly impacts mortality risk estimation based on these comorbidities.