Abstract
Recent studies reported the intraluminal thrombus (ILT) among others strongly affects abdominal aortic aneurysms (AAA) expansion rates. Thus, we investigate characteristics of ILT with AAA expansion. We applied homogeneous multistate continuous-time Markov chain models to longitudinal data of 26 Korean AAA patients as a retrospective clinical study. We considered four states of AAA and maximal thickness of ILT ($max_{ILT}$), fraction of wall area covered by ILT ($area_{frac}$) and fraction of ILT volume ($vol_{frac}$) as possible covariates. Based on likelihood-ratio statistics, $area_{frac}$ is the most significant biomarker and $max_{ILT}$ is the second most significant. Besides, within AAAs that developed an ILT layer, we found that an AAA expands relatively fast at an early stage but the rate goes slower once AAA reaches in a large size. Results recommends surgical intervention when any patient with $area_{frac}$ more than 60% or $max_{ILT}$ more than 30mm. Although this recommendation should be considered with caution given the limited sample size, one can use the proposed model as a tool to find such recommendations with their own data.
Introduction
Abdominal Aortic Aneurysm (AAA) is a potentially lethal condition characterized by the dilation of the aorta. If a rupture occurs, the mortality rate is extremely high. Current clinical guidelines often rely on a diameter threshold (typically 5.5 cm) to recommend surgical intervention. However, this "one-size-fits-all" approach is imperfect, as some small aneurysms rupture while larger ones remain stable.
The presence of Intraluminal Thrombus (ILT)—blood clots within the aneurysm—is a known factor influencing mechanical stress on the aortic wall. This study aims to quantify how specific geometric features of ILT, such as thickness and surface area coverage, affect the progression of AAA through different stages of severity.
Methods
The researchers employed a homogeneous multistate continuous-time Markov chain (CTMC) model to analyze longitudinal data. This approach is particularly suited for clinical data where patient follow-ups occur at irregular intervals.
Disease States and Covariates
The model categorized AAA progression into four states based on diameter ($D$):
- Early: $D < 40$ mm
- Mild: $40 \le D < 47$ mm
- Severe: $47 \le D < 58$ mm
- Fatal: $D \ge 58$ mm
Three ILT metrics were evaluated as covariates to predict transitions between these states: maximal thickness ($max_{ILT}$), fraction of wall area covered ($area_{frac}$), and fraction of total volume ($vol_{frac}$).
Dataset and Longitudinal Analysis
The study analyzed retrospective data from 26 Korean AAA patients. Each patient underwent multiple surveillance CT scans (up to 7) over several years, providing a detailed longitudinal view of aneurysm growth.
As shown in the longitudinal data, AAA expansion often follows a non-linear path. A key finding was the "early sprint" phenomenon: AAAs tend to expand relatively quickly in the early stages, with the growth rate slowing as the aneurysm reaches a larger size. This may be due to the mechanical cushioning effect provided by significant ILT accumulation in later stages.
Results
The statistical analysis identified the fraction of the aortic wall covered by the clot ($area_{frac}$) as the most critical biomarker for predicting disease progression.
As indicated in the table above, $area_{frac}$ yielded the highest likelihood-ratio statistic (23.33, $p < 0.0001$), followed by $max_{ILT}$. This suggests that the extent to which a clot "grips" the aortic wall is more mechanically significant than its total volume. Based on these findings, the authors suggest considering surgical intervention if $area_{frac}$ exceeds 60% or $max_{ILT}$ exceeds 30 mm.
Conclusion
This study demonstrates that ILT characteristics, particularly the wall area coverage, are vital predictors of AAA progression. While the sample size of 26 patients is a limitation, the multistate Markov model provides a robust framework for understanding the "silent" progression of aneurysms. Future research with larger datasets will further refine these predictive models and help personalize surgical recommendations.
Summary updated