Aortic aneurysms (AAs) encompass a group of pathologies affecting the aorta in any of its segments from the ascending aorta to the infrarenal abdominal aorta, which is the most prevalent type [1]. Apart from true degenerative aneurysms, dissections are commonly included in this group of conditions or summarized as aortic syndromes due to a similar enlargement of diameter and treatment options [2]. The aorta is now considered an organ rather than a simple conduit to other tissues due to its complex structure and function [3]. Considering the high rate of adverse outcomes during acute ruptures, preemptive treatment is favored when possible in elective settings and is associated with significantly lower rates of perioperative morbidity and mortality [4]. Over the past 2 decades, shifts in global cardiovascular risk factors, the establishment of national screening programs, and the widespread adoption of endovascular aneurysm repair have significantly altered the epidemiology and management of AA [5-7]. Most high-income countries have experienced a marked decline in both incidence and mortality rates for elective and ruptured AA. However, demographic changes and expanded treatment options for older and frail patients have introduced new challenges [5,8].

Estimated life expectancy is frequently considered to be a relevant factor in clinical decision-making as the risks and benefits of treatment options may be assessed differently for patients with varying life expectancies. It is broadly suggested that younger and healthier patients can reasonably be exposed to the higher perioperative risks of open surgical repair to balance out limitations of endovascular strategies including concerns with durability and imaging surveillance [4,9]. The tradeoff between short-term survival advantages of endovascular repair and potential long-term disadvantages is not yet fully elucidated and, in general, is more nuanced and influenced by patient characteristics, expectations, anatomy, and comorbidities [10].

Analyses of large population-based datasets and administrative registries have identified several risk factors associated with adverse perioperative outcomes and impaired long-term survival. However, predicting mortality and serious adverse events is complex, and traditional association studies alone may not be sufficient [11-14].

Unsupervised clustering algorithms can be used to identify patterns within complex clinical data, enabling a different approach to risk stratification by grouping patients with similar phenotypes. Unlike traditional regression models, which isolate independent effects and may produce high hazard ratios for rare but extreme risks, cluster analysis can identify clinically coherent patient archetypes where multiple low and moderate risk factors can accumulate to define high-risk subgroups. This is particularly valuable in surgical cohorts, where multimorbidity, rather than selective patient factors, frequently determines outcomes.

In this work, we investigated whether an unsupervised cluster analysis based on demographic and clinical patient characteristics and naïve to postoperative outcomes can be leveraged to identify distinct postoperative survival patterns for specific patient groups.

The study analyzed 1801 individuals undergoing open surgical or endovascular repair for AAs. The majority were male (82.3%) with a median age of 69 years. Cardiovascular comorbidities and smoking were highly prevalent. AAs were primarily abdominal (66.9%). Overall, 29.5% of patients did not survive the observational period. Unsupervised cluster analysis revealed distinct groupings exclusively based on age. In a 2-cluster model, the clusters were separated at a median age of 62 and 73 years. A 3-cluster model further stratified the population, with median ages of 57, 68, and 75 years. Survival analyses highlighted lower postoperative survival in older clusters relative to treatment, with perioperative mortality disproportionately impacting these clusters. There was strong evidence that age was significantly associated with all-cause mortality overall but showed diminished significance in older clusters after excluding perioperative deaths. Adjusting for confounders confirmed this observation, with age losing significance in the youngest and oldest clusters in later analyses.

That clusters were primarily fitted based on age is not necessarily intuitive from a physician’s perspective. We would have expected the presence of ruptures and the specific type of pathology to exert a larger influence on cluster formation, as these factors all have a relevant influence on clinical practice. Interestingly, the strictly data scientific approach of the analysis attributed heterogeneity between the included patients predominantly to age at the time of surgery. We hypothesize that a combination of collinearity, for example, a higher rate of dissections on younger patients, and homogeneity, with similarly distributed comorbidities, could explain this finding. Furthermore, aortic surgery, which is by definition a high-risk procedure, may homogenize the investigated patient sample (at least for the electively treated patients) as patients must prove a certain level of health to qualify for such an operation. Age is strongly associated with postoperative mortality following invasive treatment of aortic pathologies, but there is solid evidence that this relationship varies across age groups. In the older clusters, the diminished association between age and all-cause mortality after excluding perioperative deaths suggests that perioperative risk plays a critical role in the age-related mortality disparity. Specifically, older patients demonstrated disproportionately higher all-cause mortality within the first 30 days after surgery, implying that the perioperative period is a key driver of mortality in this subgroup. This complies with previous evidence that reported a perioperative survival disadvantage after invasive aneurysm treatment in older patients [27,28].

This differential impact across age groups highlights the complexity of the relationship between age and mortality risk, which does not appear to be linear. While overall mortality increases with age, the cluster analysis revealed thresholds or transition zones in risk rather than a simple uniform progression. The higher perioperative mortality in older patients might reflect their greater vulnerability to surgical stress, emphasizing the importance of targeted perioperative management in this group. Additionally, the lack of a strong age-mortality association in the older clusters after the perioperative period suggests that factors beyond age, such as the type of aneurysm or co-existing conditions, may have a greater influence on long-term survival in these populations. Overall, the data support a nuanced interpretation of age as a risk factor, driven largely by perioperative mortality. Investigating nonlinear relationships between age and outcomes such as mortality in biomedical research has been of growing interest in the previous years [29-32]. This work contributes to the cumulative body of evidence that might gradually induce a paradigm shift in patient risk stratification.

From a statistical perspective, the findings reflect a nonlinear association in which the hazard function for age is dominated by early postoperative events in older patients, leading to attenuation of the age-mortality relationship once perioperative deaths are removed. This pattern is consistent with threshold-like behavior rather than a single global linear term, suggesting that conventional models assuming proportional and constant age effects may obscure important variation across the age distribution.

The clinical implications of this work emphasize the need for tailored perioperative management strategies that account for age-related differences in mortality risk. The necessity to identify solutions for an aging population is not confined to the field of healthcare, but is a subject of considerable interest within it. It is estimated that by 2050, 129.8 million individuals in Europe will be aged 65 years or older [33], with significant implications for the demand on national health care systems and the surgical workforce [34-36]. Therefore, the question of age as a limitation for operative procedures such as AA repair has attracted increasing attention in vascular surgery [27] but is discussed in other fields such as orthopedics with knee arthroplasty [37] as well. A substantial body of research has indicated an elevated risk of perioperative complications for older patients across various fields, including common abdominal [38-40] and vascular procedures [41,42]. In addition to adverse outcomes, surgical treatment in older patients has been demonstrated to be more resource-intensive, characterized by prolonged hospital stays [35,43]. The unsupervised cluster analysis identified distinct age thresholds at surgery, providing a framework for stratifying patients into risk groups. This approach moves beyond arbitrarily defined age-based subgroups, such as octogenarians or nonagenarians, which have traditionally been used as proxies for elderly populations in surgical research and have been previously questioned regarding their use [44].

However, the sensitivity analysis showed that alternative age cutoffs at 65 and 75 years produced similar Kaplan-Meier survival curves, suggesting that the transitions between age-based clusters are fluid rather than strictly defined. Consequently, these thresholds should not be interpreted as rigid cutoffs but rather as approximate transition zones that may guide clinical decision-making. For example, patients in the older age groups are at increased risk for perioperative mortality, highlighting the need for enhanced perioperative care, such as optimization of cardiovascular and respiratory function, and close monitoring during the immediate postoperative period. Similarly, younger patients may benefit from strategies aimed at mitigating long-term risk factors, such as smoking cessation and management of hypertension.

The identification of fluid but meaningful thresholds by the cluster analysis provides a valuable tool for clinicians, balancing the need for structured risk stratification with the recognition of patient-specific variability. From a group-level perspective, comorbidities were rather homogenously distributed among clusters. This may be due to case selection, as the analysis included only patients deemed suitable for surgery. As a result, eligibility for surgical treatment could have a homogenizing effect on this patient sample. Nevertheless, on an individual scale, this is certainly not the case and also requires attention when evaluating invasive treatment for a given patient. Age at surgery can give some guidance when evaluating perioperative and postoperative risks, and they vary for distinct age groups, but other factors still impose meaningful effects on survival as well.

The reliance on observational data introduces the potential for selection bias, particularly evident in the older clusters, where longer median survival age after surgery may simply reflect the fact that these patients underwent surgery at a later age. This delayed treatment may be indicative of slower disease progression, possibly due to reduced exposure to risk factors such as smoking or poorly controlled hypertension, but it complicates direct comparisons between age groups. Additionally, the UK Biobank lacks detailed information on aneurysm morphology, including diameters and growth rates, which are critical for understanding disease severity and progression. Such missing data may limit the granularity of risk stratification and affect the interpretation of clinical outcomes. Data from the UK Biobank are also subject to a volunteer selection bias. Furthermore, the use of administrative and diagnostic coding introduces the possibility of errors or misclassification in identifying aneurysm type, surgical interventions, and comorbid conditions. Future studies incorporating more comprehensive clinical data and prospective designs could address these limitations and further refine the understanding of age-related risks in aortic aneurysm surgery. This will also address the need for further external validation of these findings.

This study highlights age as one of the key factors in postoperative all-cause mortality following AA surgery, with older patients experiencing higher perioperative mortality. Unsupervised cluster analysis revealed age thresholds that guide risk stratification reflecting transitions in age-related mortality risk.