Ruben Buijs

The application of computed tomography (CT) biometry in abdominal aortic disease is aimed mainly at achieving a reliable risk assessment of unfavorable outcome as early as possible. The studies presented in this thesis were centered on CT imaging of the abdominal aorta, especially in context of aneurysmal disease and subsequent potential endoleak after endovascular repair. The current consensus on abdominal aortic aneurysm (AAA) repair is centered on its most important biometric variable, the maximum aneurysm diameter. Despite the AAA diameter being the strongest predictor for rupture risk available, there is a need for reinforcing its predictive value with additional risk factors. Aneurysmal weakening of the aortic wall follows a similar pathophysiological path as vascular calcification does, is easily assessable with CT imaging and has been continuously correlated with adverse events in the cardiovascular system. This prompted CT-based calcification scoring to be identified as a potential prognostic biometric factor for AAA outcome. In the treatment of AAA, endovascular aortic repair (EVAR) has been established as an interventional option that performs as well as, and in certain circumstances better than, open surgical aortic repair. The endograft type and size are chosen based on the biometry of the aneurysm as provided by CT angiography. Faulty sizing can lead to inadequate sealing of the endograft at the proximal aneurysm neck, leading to type 1A endoleak. Endoleakage is a major contributor for re-intervention and conversion to open repair. There are different endograft sizing techniques, although none have undergone peer-reviewed qualitative analysis or comparative assessment between each other.

For this thesis, the aim was to investigate novel developments in prognostic biometrics in computed tomography in patients with abdominal aortic aneurysm. First, chapter 2.1 provided a review of contemporary imaging modalities and their applications towards improving rupture risk assessment of AAA. A multitude of imaging modalities were found, with varying applications of ultrasound (US), magnetic resonance (MR), positron emission tomography (PET), CT imaging, as well as bio-optical imaging. Some studies provided novel approaches to AAA risk assessment with a good chance of clinical implementation, such as computational CT angiography analysis and 18-F PET-imaging. Studies in the field of experimental US, MR and bio-optical imaging modalities were mainly performed with small cohorts and were all far from clinical application. Results garnered in these studies generally lacked scientific strength, caused by low population sizes, disagreement between studies and uncertain clinical relevance, as most were far from implementation in routine care. Nonetheless, calcification analysis was reported to be of value, both in terms of ease of visualization and clinical significance in both chapters 2.1 and 2.2. Chapter 2.2 also reported on the contradictory results of CT-based calcification scoring of AAA patients. Most authors hypothesize that calcification is a proxy of vessel wall disintegration. Higher degrees of calcification were correlated with decreased vascular elasticity and compliance, as well as with greater cardiovascular morbidity and mortality. On the other hand, calcification theoretically provides mechanical protection against the shear wall stress of passing blood or is a steady state of a different, gradual form of atherosclerosis, as opposed to a more disintegrative form of atherosclerosis. The controversy is exemplified in the concluding remarks posed by Lindholt et al., stating that calcified AAA are more likely to follow the natural course of small AAA, as opposed to having greater risk of rupture. There is little further evidence in agreement of a negative relation between aortic calcification and risk of AAA rupture. So, despite the ambiguous role of calcification in aortic disease and CT imaging thereof, most available research points towards a prognostic role for aortic calcification in AAA rupture risk assessment.

AAA calcification has been investigated to some degree, mainly with regard to non-aneurysm related disease outcomes. Yet, it has not been studied as a risk factor for AAA rupture. Chapter 3.1 starts with a clinical investigation of aortic calcification on CT imaging in a retrospective unmatched case-control population. The aim of this study was to assess whether aneurysm calcification is correlated to rupture. At first glance, the results support the predictive value of calcification of the abdominal aorta on CT. In accordance with common knowledge, symptomatic and ruptured AAA patient groups showed greater aneurysm diameters. Also, greater aneurysm calcification scores were found in these groups in comparison to the control group, electively treated AAA patients. However, chapter 3.1 concludes with a decision not to overestimate the role of aortic calcification, as there were significant limitations to the methodology. The retrospective nature of the study allows for more selection bias to occur, and it negates any possibility to attribute causality in its analysis. Most importantly, the applied AAC-8 score is limited by its rough assessment of aortic calcification. Essentially, it applies a binary grading system, divided in four segments of the abdominal aorta, which provides a 0 to 8 grading scale. This is a far cry from fully quantitative measurements of calcium mass, specified up to the milligram. However, in practice is does provide clear categorization of vascular calcification grade, in a similar manner as the Agatston score has done for decades in CT imaging of coronary artery calcification. Also, in contrast to other calcification scoring tools that were available at the time, the AAC-8 score had been applied in multiple publications, it is easy to use, and it is applicable to contrast-enhanced CT angiography images. Manual segmentation of intravascular contrast allowed for detailed visualization of aortic calcification in all scans. However, this does emphasize the rater-dependency of the tool, another clear limitation of the AAC-8 score. The inter-rater reliability was good, but a fully automated calcification score would bypass observer-bias completely and therefore provide a greater degree of reproducibility and standardization.

The results of chapter 3.1 warrant further investigation, though before continuing an investigation of aortic calcification in relation to AAA rupture risk on a larger scale in a prospective cohort, any calcification measurement tool should at least have undergone some form of qualitative appraisal. Of the clinical studies that have applied aortic aneurysm calcification scoring in some form, few have provided references to peer-reviewed studies that support the measurement tool. Even fewer of these studies provide a rigorous qualitative appraisal of the measurement tools under the circumstances that these will be applied in. Chapter 3.2 was performed with the aim to do just this. The study assessed the effects of CT acquisition parameters and intravascular iodine contrast on the measurements of aortic calcification on CT images. Results showed that calcification volume and mass was overestimated to an extreme extent under all applied scanning circumstances. This was combined with a wide variance in overestimation, impeding the potential use of a correction algorithm. Reliability was reduced further by the presence of iodine contrast. Yet, intravascular contrast is inextricable for the diagnostics and pre-operative sizing of AAA, so performing experimental AAA calcification scoring of non-contrast enhanced CT images is of little value for clinical practice. Especially since there is no clinical foundation to warrant the radiation exposure of an additional non-contrast enhanced CT scan. Dual-energy CT may eventually pose as a practical alternative in this regard (chapter 2.1). The results of chapter 3.2 are corroborated by earlier work on a different calcium scoring tool, by Komen et al. Scoring outcomes differed significantly with changes in CT slice thickness, lower Hounsfield Unit thresholds and the presence of intravascular contrast. Changing the convolution kernel, a significant component of standardized CT scanning protocols, did not affect calcium scores. Our research expands on their findings by comparing the coronary versus the abdominal CT protocol, by maintaining the same slice thickness, while changing the convolution kernel and the amount of milliampere seconds (mAs). No relevant differences in outcomes were found comparing the CT imaging protocols. Also, chapter 3.2 discusses that there are currently multiple types of calcification scoring tools available, most of which have no scientific basis for the level of reliability that is placed upon them. Because of this, a gold standard of AAA calcium scoring tools is yet to be established. Chapter 3.2 applies an experimental scoring tool that has not been tested for AAA calcification in any peer-reviewed scientific publications. Nonetheless, the 3mensio Structural Heart scoring software works according to the same procedure as any other fully quantitative scoring tool. Theoretically, the most accurate scoring is performed by mass or volume measurement tools that provide a result in, for example milligrams or mm3. One such tool was applied in chapter 3.2. The alternative is semi-quantitative and qualitative scoring. The most accurate example of this is the Agatston score which categorizes the degree of calcification within certain ranges. Calcification can be expressed as a percentage of calcified aortic wall in areas or circumferential segments, or even less specific, by qualifying the presence of calcification per aortic segment in binary terms (present/absent), such as in the AAC-8 or AAC-24 score (detailed in studies of chapter 2). There is a host of studies that aimed to correlate clinical outcomes to the degree of aortic calcification, by using the abovementioned scoring methods or slight variations thereof, yet only one publication by Komen et al. has previously evaluated the effect of a set of common clinical CT imaging variations on the scoring results of their measurement tool, the Siemens Calcium Score. The fact that currently only two peer-reviewed studies have studied some components of reliability of two different tools for AAA calcium scoring, is worrisome. Especially given the fact that most of the before mentioned studies applying AAA calcium scoring tools do not employ either of these tools. It is worrying because there is no scientific basis for the assumption that the scoring outcome of any employed AAA calcium scoring tool is reproducible, and therefore its reliability is unknown.

Prognostic biometry is also applicable to perioperative CT imaging, for the prevention of important surgical morbidity, such as type 1A endoleak (EL1A). The articles in chapter 4 provide an approach to the prevention of EL1A. Firstly, in chapter 4.1, a consensus update is provided on the prevention of EL1A in endovascular aortic repair (EVAR) in comparison with endovascular aneurysm sealing (EVAS). Inadequate endograft sizing is one of two risk factors for EL1A that are dependent on the skill and experience of each individual vascular surgeon, the other one being perioperative manual surgical prowess. For endograft sizing, reducing the dependency on the individual surgeon’s skill in terms of inter-rater variability should hypothetically reduce the incidence of EL1A. The current standard of endograft sizing is dependent on the diameter of the proximal aortic neck. This is based on the mean of the largest and smallest axis of the neck and therefore has limited use in very short or tortuous aortic necks. In routine clinical experience these anatomical configurations are known as “hostile”, as these are correlated to a higher incidence of adverse postoperative effects. Under these circumstances, it is theorized that endografts will be inaccurately sized to be smaller than their true diameter, potentially impeding the improved sealing effect of oversizing. Chapter 4.2 essentially provides a mathematical approach to the hypothesis that the aortic neck circumference is a mathematically correct reference for the calculation of the diameter and concomitant endograft sizing. Not only does this study provide a theoretical basis for clinical investigation of the circumference-based endograft sizing method, it is also a mathematical assessment and critique of the traditional method. It proposes that the traditional method will result in greater underestimation of the true aortic neck diameter in tandem with increased differences between the two axes that make up the mean diameter-based method. Chapter 4.3 provides follow-up of this hypothesis, by applying the circumference-based method in a retrospective clinical cohort, and by comparing it to the traditional mean diameter-based method. The main result of this comparison yielded no relevant distinction between the novel circumference-based method and the traditional method. Neither were there differences between the type 1A endoleak (EL1A) patient group and the control group, for either method. The population for this study was too small to attain adequate statistical power, so the statistical outcomes of the study could by definition not be extrapolated to the general population. Fortunately, this study yielded more information on a case-by-case basis and may eventually contribute to this scientific field at a meta-analytical level, especially regarding the rarely studied EL1A population. For instance, endografts of eight out of 12 EL1A cases had been inadequately oversized. Oversizing may therefore not only function to exert greater radial force on the aortic neck, thus improving the sealing of the prosthesis. It may have also become a means to negate the lacking accuracy of the current gold standard for endograft sizing. Perhaps it is in accordance to the adage “do not fix something that is not broken”, that no improvements to the traditional method have been previously tested, prior to this thesis. Moreover, with the development of surgical alternatives such as EVAS, the search for such improvements will be unnecessary. However, as the inter-rater variation was sizeable in chapter 4.3, endograft measurements may be undersized by a significant margin. Considering the fact that optimal oversizing is between 10-20% of the measured dimensions, the inter-rater variation may play a role in excessive oversizing or undersizing in some patients. Moreover, EL1A after EVAR continues to occur at incidences averaging between 3.3 and 20.1%. A study by van Marrewijk showed that within a 2-year period post-EVAR, 59% of patients with EL1A and the less common type 3 endoleak underwent secondary interventions such as additional ballooning, aortic cuff placement, or conversion to open repair (10.8%). Patients without endoleak showed a 9% rate of secondary intervention, with 0.8% undergoing conversion to open repair. In a recent study O’Donnell et al. state that of all EL1A patients (8%), none were free from secondary interventions, and 11% underwent one or multiple incidences of open conversion or endovascular reintervention. It is uncertain to what extent inadequate sizing was causative in these instances, but according to the results posed in chapter 4.3, the extent may be significant and should at least warrant additional research.

There is a wide range of potential risk factors and experimental imaging modalities for AAA risk assessment. While most lack in terms of clinical adaptability, calcification scoring is a novel biometric for CT angiography that is both promising and fairly simple to adopt. Vascular calcification of the aorta is mostly regarded as a proxy of vessel wall disintegration. In spite of some evidence in favor of a protective role for aortic calcification against AAA rupture, most contemporary research suggests a prognostic role for CT-based calcification scoring in AAA rupture risk assessment.

The present study suggests that calcification of the abdominal aorta might have predictive value in AAA rupture risk assessment. Also, as opposed to AAA diameter, calcification scoring appeared to discriminate symptomatic aneurysm patients from those that underwent elective repair. Among other important limitations, the AAC-8 calcification scoring tool has suboptimal accuracy and remains observer-dependent. An automated, fully quantitative software tool should be able to improve on either. In search of such a tool, we found that AAA calcification has often been correlated to non-aneurysm related disease outcomes, with a varying array of CT-based calcification scoring tools. Very little scientific evidence has been provided to back-up claims of reliable use of these tools. This thesis provides the second ever technical assessment of a CT-based calcification tool for AAA. Besides confirming parts of previous research by other authors, this study found overall gross and incorrigible overestimations of calcification volume and mass under all applied scanning circumstances. Intravascular iodine contrast further disrupted reliable calcification scoring by a wide margin. If extrapolated to other Hounsfield unit-based automated calcification scoring tools for CT angiography, these results suggest that many previous studies applying similar scoring tools, ought to be scrutinized.

Lastly, prognostic biometry can also be applied for prevention of type 1A endoleak (EL1A) in the preoperative-phase. Endovascular aortic sealing may potentially decrease the incidence of EL1A. However, after endovascular aortic repair, EL1A remains a major cause for morbidity and reinterventions. Endograft sizing and adequate oversizing is important in the prevention of EL1A, specifically in patients with hostile aortic neck characteristics. Despite this fact, there have not been any scientific publications on the reliability of endograft sizing methods, nor has there been comparison against alternatives. In this thesis, a novel, circumference-based endograft sizing method is shown to be mathematically more accurate than the traditional mean-diameter based method. Nonetheless, there was no clinically measureable difference between either method. The inter-rater reliability of both the novel and traditional method was low and may have led to undersizing and even extreme oversizing in some EL1A cases in the studied cohort. Given these results and overall lack of scientific research in the field of endograft sizing, the widespread reliance on the traditional method is not fully justifiable and requires further research.