PhD Candidate McGill University Montréal, Quebec, Canada
Background: Current guidelines for the management of aortic disease describe different patient groups at risk. In this study, a machine learning approach was used to better understand our Aortic Clinic patient population groups using clinical variables and compare the group biomechanics.
METHODS AND RESULTS: A cohort of 190 patients having undergone aortic resection surgery in our clinic between 2018 and 2023 for ascending aortic aneurysm repair was investigated. Patient characteristics (basic demographics, genetics, aortic dimensions, etc.) were fed to an unsupervised machine learning algorithm (k-cluster analysis) to stratify our population. This analysis created a series of clinically significant patient groupings. Biaxial mechanical testing was performed on the resected aortic tissue, and mechanical properties were compared between groups.
Four relevant patient subgroups were observed using this approach (Figure 1). Three of these groups feature patients that largely follow the American Heart Association guidelines for aortic disease management. The first group contains more patients with bicuspid aortic valves, aortic stenosis and smaller aortic diameter (P < 0.01). The second group features older patients with more risk factors, and high aortic diameter-height ratio (P < 0.01). The third group was younger with few risk factors while being tall with syndromic featuring (P < 0.01) along with higher incidence of positive genetic testing (P < 0.05). A fourth group features patients with higher incidence of severe aortic root dilation, aortic insufficiency and negative genetics (P < 0.01). The second group (older patients, high aortic diameter/height) had the poorest mechanical properties when assessing energy loss (P < 0.0001) while there was no significant difference in aortic tissue thickness between the groups. Normalizing by age shows the second group having significantly higher energy loss (P < 0.005 with Cluster 2 and 4, P< 0.0001 with Cluster 1), with the rest of the groups being comparable to each other. Histological analysis revealed higher collagen-elastin ratio in the second group compared to the first and fourth groups (P < 0.05).
Conclusion: Current guidelines used for surgical recommendation for aortic aneurysms have criteria that can depend on patient phenotype. Differences between the subgroups obtained from cluster analysis broadly match existing guidelines. The fourth group, however, does not fit into any one guideline grouping, as it contains patients who have large root aneurysms without genetic positives, as well as being comparable to non-syndromic groups when accounting for age.
