Aortic elongation and the risk for dissection: the Tübingen Aortic Pathoanatomy (TAIPAN) project†

Abstract

INTRODUCTION

The ascending aorta diameter is the only established morphological risk factor for Stanford type A aortic dissection (TAD), and it triggers prophylactic surgery [1, 2]. However, it has been demonstrated that the majority of TAD develop below the threshold diameter of 55 mm [3, 4]. Therefore, screening for patients at risk for TAD is ineffective.

Recently, we hypothesized that aortic elongation may be another risk factor for TAD, and it could be useful for screening [3]. Elongation, accompanied by changes in the longitudinal mechanical properties of the aorta [5, 6] may be an aspect of the pathogenesis of TAD. We showed that the ascending aortic projections of patients pre- and post-TAD are elongated compared to healthy controls in frontal and sagittal 2D computer tomography angiography (CTA) reconstructions. However, so far, we have not assessed the 3D anatomic length of the aorta [3].

In this study, we assessed aortic morphology in pre- and post-TAD patients and in a healthy control group, using contemporary 3D-CTA reconstruction techniques. This enabled us to study not only realistic diameters but also anatomical length values of the different aortic segments. The aim of this project was to refine the risk-morphology for TAD, analyse the prognostic value of aortic parameters and develop a multidimensional score that improves TAD prophylaxis.

METHODS

Patient groups and clinical data

We retrospectively compared 3 groups of patients. We analysed all patients treated for TAD at our centre between January 2006 and December 2015 (n = 172). We excluded those being younger than 40 years of age (n = 3), and none of the analysed patients had documented connective tissue diseases. Within this cohort, we identified patients who had received an adequate CTA within 24 months before the actual dissection occurred (4 patients with CTA performed more than 2 years prior to TAD were excluded). Those patients (n = 15) formed the preTAD group and the remaining 150 patients formed the TAD group (statistically independent). In the preTAD group, the median time between preTAD CTA and the actual dissection was 3.8 months (Q1–Q3: 1.7–8.4; range: 0.6–21.7 months). We also screened all patients diagnosed in our emergency department with an adequate CTA between March 2014 and December 2015 because of non-aortic emergencies (n = 247). Expectedly, these included very young adults. For homogenization, we excluded all patients below age of 40 from this emergency department group (n = 32). This resulted in a healthy control group of 215 patients with similar demographics compared with the TAD group.

We recorded the following demographical and clinical parameters: date of birth, date of the CTA, height, weight and sex. We screened the medical files for a diagnosis of hypertension and the presence of the following drugs: beta blockers, angiotensin-converting enzyme (ACE) inhibitors, AT-II-receptor antagonists, vasodilators and calcium channel blockers, but not diuretics. The presence of 3 or more of these drugs in the chronic medications indicated patients suffering from massive hypertension.

This study was approved by our local ethics commission (No. 076/2015R). Obtaining written informed consent was not necessary because of the retrospective observational character of the study.

Computed tomography

CTA studies were performed using a 2nd generation dual-source CT scanner (Somatom Definition Flash, Siemens Healthcare, Erlangen, Germany). A non-ionic, high-iodinated contrast bolus tailored to body habitus (∼100 ml; 400 mg/ml iodine) was injected at a high-flow rate (>4–5 ml/s), followed by a saline chaser. In the majority of the cases, an automated bolus triggering a region of interest located in the ascending aorta was used. We included subjects in whom CTA images with a maximum thickness of 3 mm were available.

Image processing and analysis

CTA datasets were processed using the OSIRIX-MD (PIXMEO, Bernex, Switzerland) PACS-Viewer and image-processing software package. Curved multiplanar reformats were used to visualize the aorta. In detail (Fig. 1), a curved multiplanar reformat was produced by manually defining the aortic central line with a 3D-Bezier-path in the frontal, sagittal and transversal CTA-reconstructions, optionally by angulating the frontal and sagittal planes according to the direction of the aorta. At defined landmarks, orthogonal to this centreline, cross-sectional reconstructions were produced (‘true short-axis’ images) [7–9]. The definition of the landmarks adhered to international guidelines [1, 2, 7]. However, slight modifications had to be made because some landmarks in the guidelines were difficult to reproduce (i.e. the diaphragm) and were replaced by distinct landmarks (i.e. the celiac trunk):

Figure 1:

(A) Aortic landmarks. (B) Sagittal, frontal and transversal reconstruction of the CTA. (C) Curved multiplanar reformat of the entire aorta. (D) True short axis views at the landmarks.

Open in new tabDownload slide

(D1) aortic valve annulus (AV).

(D2) sinus of Valsalva (halfway between D1 and D3).

(D3) sinotubular junction (STJ).

(D4) mid-ascending aorta (halfway between D3 and D5).

(D5) orifice of the brachiocephalic trunk (BCT).

(D6) mid-aortic arch (halfway between D5 and D7).

(D7) distal aortic arch (directly downstream the left subclavian artery).

(D8) descending aorta (at the height of the pulmonary artery bifurcation).

(D9) thoraco-abdominal (at the orifice of the celiac trunk).

(D10) mid-abdominal (halfway between D9 and D11).

(D11) distal abdominal (directly proximal the aortic bifurcation).

The aortic segments and length values were defined as follows, also corresponding to predefined conventions [7]:

(L1) aortic root (AV to STJ, D1 to D3).

(L2) ascending aorta (STJ to BCT, D3 to D5).

(L3) aortic arch (BCT to distal of subclavian artery, D5 to D7).

(L4) distal aortic arch (subclavian to pulmonary artery bifurcation, D7 to D8).

(L5) descending aorta (pulmonary artery bifurcation to celiac trunk, D8 to D9).

(L6) abdominal aorta (celiac trunk to bifurcation, D9 to D11).

The length parameters (L1–L6) were measured along the centreline in the curved multiplanar reformats after identifying the mentioned landmarks (D1–D11). To minimize failure resulting from non-circularly shaped aortas and from the non-exact placement of length-measuring tools, we measured the aortic perimeter and calculated the mean derived diameter. We placed the measuring tool within the aortic wall, retracing the shape of the aorta including the true and the false lumen. If thrombus was present, it was included in the measurement.

We also recorded the morphology of the aortic arch in the sagittal plane; in the type I arch configuration, the highest and reversal point of the arch is between the supra-aortic vessels and in the type II arch configuration, this point is distal of the left subclavian artery [10, 11].

Statistical analysis

Continuous data were checked for normality by analysing skewness and kurtosis and by performing the Shapiro–Wilk test. Because most, but not all, variables fulfilled the criteria for normality we waived the presentation of parametrical statistics. Continuous data were described with the median and the first (Q1) and third (Q3) quartile, and the range (minimum and maximum) and are presented with box-and-whisker plots. Categorical data are reported as percentages. Spearman's rank correlation coefficient was used to determine the correlation between continuous variables. Analogously, we performed point and point-biserial correlations. For inferential statistical comparisons of continuous variables between the 3 study groups, we performed a closed testing procedure: we first ran a Kruskal–Wallis test. In the case of a significant result, we secondarily performed pairwise Mann–Whitney–Wilcoxon rank sum tests. Analogously, categorical variables were compared using χ² tests (⁠$3\hspace{0.17em}\times \hspace{0.17em}2$— followed by $2\hspace{0.17em}\times \hspace{0.17em}2$ tests). All reported P-values were two-sided and P-values of ≤0.05 were considered to indicate statistical significance.

Receiver operating characteristic (ROC) curves were plotted and analysed for the healthy control group and the preTAD group (not the TAD group) to assess the diagnostic value of the individual aortic diameters and length parameters [12], and to identify those cut-off values which can best differentiate between control and preTAD group. Sensitivity and specificity of the variables and scores were calculated for the preTAD and control groups based on the resulting contingency tables.

SSPS 23.0 (IBM Corp., Amonk, NY, USA) and MS Excel 2010 (Microsoft, Redmond, WA, USA) were used for analyses and data presentation. Statistical analysis was performed in accordance with our association's guidelines [13].

RESULTS

Aortic aging

To investigate the physiologic age- and body size-dependent changes in aortic morphology, we analysed the aortic dimensions in all patients who received a CTA in our emergency department (including those <40 years).

The ascending aorta diameter (D4) was moderately correlated with age (r = 0.509; P < 0.001), whereas correlations with body height, weight or body mass index (r = 0.05, P = 0.42; r = 0.23, P < 0.001 and r = 0.24, P < 0.001; respectively) were insubstantial. This moderate positive correlation with age was also observed for other diameters of the aorta, namely at the BCT (D5; r = 0.523; P < 0.001) and the arch (D6; r = 0.519; P < 0.001). No significant correlations were found between age and root diameter or descending aorta diameter, also no significant correlations were found between aortic diameters and bodyweight and height.

With respect to aortic length, the distal arch length (L4) was positively correlated with patient age (r = 0.636; P < 0.001). In the ascending (L2) and arch (L3) sections, these coefficients were clearly smaller (r = 0.331 and r = 0.326, both P < 0.001 respectively). The root (L1), and descending lengths (L5) did not show a significant correlation with age, nor was the aortic length in any segment significantly correlated with other body dimensions, except the descending length was modestly correlated with body height and weight to the magnitude of r = 0.38 (P < 0.001).

Figure 2 shows the scatter plots of age and selected aortic dimensions as well as the corresponding linear regression with the suitable linear equations and 95% confidence intervals. The ascending aorta median diameter increased on average by 1.4 mm per decade but the increment in length was 2.1 mm per decade. In the case of the aortic arch, the median length increment was 7 mm per decade of life.

Figure 2:

Scatter plots of age and the selected aortic dimensions; linear regression and 95% confidence intervals.

Open in new tabDownload slide

Aortic dimensions and hypertension

In the control group, 43.3% of patients suffered from documented hypertension, with 7.0% having massive hypertension (non-additive). In both the preTAD and TAD groups, the prevalence of hypertension (86.7% and 66.0%) and massive hypertension (40.0% and 27.3%, respectively) were significantly (P < 0.001) higher.

In the control group, the presence of arterial hypertension was significantly (P < 0.001) but weakly correlated (r = 0.3) with the diameters of the ascending aorta (D4) and the arch (D5 to D7), whereas the correlations with the other aortic diameters were insignificant. With the exception of the distal arch length (L4; r = 0.253; P < 0.001), the aortic length parameters showed no significant correlation with the presence of hypertension. Particularly, no significant correlation was found between the ascending aorta length (L2) and hypertension.

Hypertensive patients, on average, were 15 years older (71.6 vs 56.6 years; r = 0.398; P < 0.001) than normotensive patients. Therefore, age and hypertension confounded each other in this observation.

Aortic dimensions in undissected and dissected aortas

Table 1 shows the demographic data of all 3 study groups. No relevant differences were found with respect to gender, age and body dimensions.

All aortic diameters (Fig. 3) in preTAD and TAD aortas were significantly (P < 0.001) larger than those of the control group. Between the preTAD and TAD groups, only the diameters representing the ascending aorta and the arch (D3–D7) differed significantly (P < 0.05), whereas the aortic root (D1–D2) and descending aorta (D7–D11) did not. The numerically largest differences in diameter were found in the mid-ascending aorta (D4). The median diameter in control aortas was 35 mm (Q1–Q3: 31–38; range: 22–47), whereas it was 43 mm (Q1–Q3: 39–46; range: 37–77) in the preTAD and 50 mm (Q1–Q3: 45–56; range: 34–94) in TAD aortas. All of the control aortas, 93% of the preTAD and 68% of the TAD aortas measured less than 55 mm in the mid-ascending aorta (D4).

Figure 3:

Aortic diameters at the landmarks in the preTAD, TAD and control groups.

Open in new tabDownload slide

Similarly, in all segments, the TAD aortas were significantly (P < 0.001) longer than the control group aortas (Fig. 4). Between the control and preTAD groups, the length differences in the ascending aorta (L2; P < 0.001), the arch (L3; P = 0.006) and the distal arch (L4; P = 0.005) reached significance, whereas those in the root (L1), the descending aorta (L5) and the abdominal aorta (L6) did not.

Figure 4:

Aortic segment lengths in the preTAD, TAD and control groups.

Open in new tabDownload slide

The most relevant length gain was found in the ascending aorta (L2). The mean length in the control group was 70 mm (Q1–Q3: 64–77; range: 40–107), whereas the same segment was significantly (P < 0.001) longer in both, the preTAD group (86 mm; Q1–Q3: 76–95; range: 55–102) and in the TAD group (92 mm; Q1–Q3: 82–99; range: 35–124).

To provide an easy way to measure the ascending aorta length and to avoid the critical identification of the STJ, we analysed the additive values for the complete ascending aorta (L1 + L2) and the complete arch (L3 + L4). In the control aortas, the median length for the complete ascending aorta (L1 + L2) was 93 mm (Q1–Q3: 85–101; range: 58–134), and it was significantly (P < 0.001) longer in both, the preTAD aortas (111 mm; Q1–Q3: 98–121; range: 76–129), and the TAD aortas (117 mm; Q1–Q3: 107–127; range: 58–150). Values for the complete arch (L3 + L4) also differed significantly (P < 0.001): Control 99 mm (Q1–Q3: 88–114; range: 52–181), preTAD 117 mm (Q1–Q3: 104–132; range: 87–166) and TAD 118 mm (Q1–Q3: 107–130; range: 70–190), respectively (Fig. 4).

Aortic arch morphology

In the control group, the highest point of the aortic arch, the reversal point, in 76.3% of patients was between the supra-aortic branches (type I arch), and in 23.7% it was downstream of the left subclavian artery (type II arch). The presence of type-II-arch morphology was weakly correlated with age (r = 0.419, P < 0.001) and arterial hypertension (r = 0.301, P < 0.001). Within the preTAD and TAD groups, 60.0% (P = 0.002) and 46.7% (P < 0.001), respectively, of patients had type II aortic arch morphology. In both groups, this constituted a significantly higher proportion compared to the control group.

Prognostic value of the variables

We performed a ROC analysis with the control and the preTAD groups to assess the prognostic value of the different variables [12]. The ROC-area under the curve (ROC–AUC) quantifies the discriminative value of an individual variable. A ROC–AUC of 0.5 indicates no discrimination, whereas 1.0 shows perfect discrimination. Table 2 shows the ROC–AUC values of selected variables; the respective values of the other variables were lower. The ascending aorta diameter (D4) showed a comparably high ROC–AUC of 0.912. A diameter of 45 mm at D4 distinguishes between control patients and preTAD patients with a sensitivity of 0.33 and a specificity of 0.98. A higher threshold; i.e. 55 mm, provides a specificity of 1.0 but identifies preTAD patients with a lower sensitivity of just 0.06. The STJ diameter (D3) showed an even higher ROC–AUC of 0.924, but the STJ is difficult to measure.

The additive length parameters complete ascending aorta (L1 + L2) and complete arch (L3 + L4) showed ROC–AUCs to a magnitude of 0.787 and 0.747, respectively. However, the variable complete arch (L3 + L4) showed an unfavourable relationship between sensitivity and specificity; at a threshold of 140 mm, the sensitivity for identifying preTAD patients in our cohort was as low as 0.07, but the specificity was still suboptimal at 0.96. At a threshold of 145 mm, this relationship was not much better. On the contrary, the parameter complete ascending (L1 + L2): a threshold of 115 mm reaches a specificity of 0.96 and a sensitivity of 0.40; a more conservative threshold of 120 mm provided a specificity of 0.98 and a sensitivity of 0.27.

DISCUSSION

Study limitations: computer tomography angiography acquisition and processing

Motion artefacts, particularly in the aortic root, may hamper exact measurement. In contemporary CTA studies, this problem may be solved using ECG-triggering or flash techniques. Defining the aortic intima in the case of suboptimal contrast media distribution is problematic and thus the luminal aortic diameter may be overestimated. We included CTA studies with an age of up to 10 years in this work. Therefore, not all studies encompassed optimal contemporary, but adequate quality. Measuring the aortic diameter with a linear measuring tool may be problematic because there is not always a strictly circular shape of the aorta; additionally, the decentral placement of the tool and the potentially non-exact differentiation of the intima-blood border can be problematic. To minimize these problems, we measured the perimeter of the aorta and calculated the mean derived aortic diameter. It is methodically somewhat difficult to identify the STJ reliably in all cases, secondary to motion artefacts and, more importantly, because it is often flattened, particularly in diseased aortas. In our opinion, parameters involving the identification of the STJ are not suitable for routine measurements; accordingly, we rely on measurement of the distance from the AV to BCT (complete ascending L1 + L2) to assess the ascending aorta length.

Study limitations: study design and analysis

The preTAD group (n = 15) was critically small, particularly considering that the ROC-analysis was based on this group. A larger cohort would enable further validation i.e. by splitting the dataset in ‘learning and confirming groups’. However, in our experience, only a very small number of TAD patients have an adequate preTAD-CTA study available. To address this problem, we are planning to collect more preTAD-CTAs in a multicentre approach.

Due to the retrospective design of this study, the ability to identify certain risk factors was limited. A prospective design would be superior but has several ethical problems, such as exposing healthy subjects to radiation and exposing patients with pathological aortic dimensions to the risk of experiencing a TAD. Furthermore, given the relatively low prevalence of TAD, a very large number of patients would have to be included to obtain reliable results. Against this background, the actual study design from our point of view was pragmatic.

Aortic aging

Changes in aortic morphology were correlated with age but not with parameters of body size, which was in accordance with the literature [14]. Concerning the ascending aorta diameter, growth rates of 0.7–0.9 mm/decade have been reported [14–16]. We actually found an even higher rate of 1.4 mm/decade. Similarly, age-associated ascending aorta elongation rates of 0.9 mm/decade have been described [14]. Again, our values (2.1 mm/decade) were higher. These differences probably resulted from different definitions and methodology. More importantly, our data showed that age-associated elongation mainly affects the arch and, to a lesser extent, the ascending aorta. This leads to the conclusion that circumferential dilatation predominantly affects the ascending aorta, whereas age-dependent aortic elongation is pronounced in the distal arch. Presupposing these dynamics, individual aortic dimensions need to be interpreted in terms of the patient's age. In this sense, the scatter plots presented in Fig. 2 may serve as a blueprint for a nomogram, whereas for the creation of valid nomograms, large, population-based cross-sectional studies would be necessary.

Pathological aortic dimensions

An effective screening for aortas at risk is certainly desirable as the perioperative mortality of elective ascending aorta procedures [2, 17] is much lower compared to emergency surgeries. The threshold diameter of prophylactic ascending aorta replacement of 55 mm is largely based on the classical publications of Elefteriades and co-workers [18, 19]. However, most dissections happen at diameters clearly below this critical diameter; Rylski et al. [4] reported a median diameter of 38.4 ± 5.5 mm at the time point of dissection. Our own data showed a median ascending diameter in the preTAD group of 43 mm. It might be objected that our preTAD group may be distorted by selection bias because larger aortas would have been operated on in the first place. However, even in the TAD group, which is certainly not biased, the median diameter was just 50 mm, and 68% of the dissected aortas had diameters <55 mm. In our analysis, the 55 mm threshold had a specificity of 100%, but the sensitivity was only 6% in discriminating between controls and the preTAD group. Decreasing the threshold to 45 mm would have led to a sensitivity of 33% and a suboptimal specificity of 98%. Basically, we cannot convey that the 55 mm threshold is chosen too high, but we must conclude that screening on the basis of the sole ascending diameter must remain ineffective. Certainly, Elefteriades et al. [18] were not wrong, as aneurysmatic aortas do have a high risk of dissection, but their prevalence in the population is low; therefore, most dissections happen at diameters <55 mm.

Aortic elongation, the longitudinal form of dilatation appears as a potential risk factor for TAD for several reasons: (i) aortic elongation is often seen clinically but poorly studied because it is not as intuitive to measure, (ii) the longitudinal compliance plays an important role in the windkessel effect of the healthy aorta [5, 6]. The wear of the longitudinal elastic properties by elongation and consecutive changes in wall architecture may promote vulnerability of the aortic wall, and (iii) the intimal tear of TAD usually runs in horizontal direction [3, 20] which suggests a trauma respectively a distinct wall vulnerability perpendicular to that, which is longitudinal.

Consequently, involvement of aortic elongation in the pathogenesis of TAD at least appears conclusive.

TAD and preTAD aortas were elongated compared to control aortas, particularly in the ascending aorta and arch. The central line distances between the STJ and the BCT (L2) in the control group were below 90 mm in 96% of patients; however, this length was exceeded in 27% of the preTAD patients and in 55% of the TAD patients. The measurement of the distance ‘complete ascending’ (L1 + L2) is way better to standardize. Within 98% of the control patients, this distance was <120 mm, whereas 27% of the preTAD and 45% of TAD aortas exceeded this value. In our opinion, this provides the basis to consider central line distances STJ–BCT (L2) of ≥90 mm and AV–BCT (L1 + L2) ≥120 mm as pathological.

Prediction of aortic dissection

A diagnostic score to recognize aortic risk morphologies should provide high sensitivity but it also must provide perfect specificity not to expose patients unnecessarily to the risk of prophylactic aortic procedures.

We propose a 2D prognostic score composed of the ascending aortic diameter and the distance from the AV to BCT for the identification of preTAD patients (Table 3). This score requires only the 3D reconstruction of the ascending aorta (Fig. 5). Patients that have at least 2 points in this score may be considered for prophylactic ascending aortic replacement. An ascending aorta diameter of 55 mm leads to 2 points and, consequently, to surgery, which matches the actual guidelines. A dilated aorta of ≥45 mm and a 3D-central line distance between the AV and the beginning of the BCT of ≥120 mm are rated at 1 point each. In our data, this score had a specificity of 1.0 and a sensitivity of 0.27, which was a quadruplication of the sensitivity of the sole threshold diameter of 55 mm. We also evaluated more complex scores involving aortic arch length and morphology. However, adding these parameters did not improve the performance of the score.

Figure 5:

Practical measurements required for the TAIPAN score.

Open in new tabDownload slide

Table 3:

Open in new tab

Parameters of the TAIPAN score

Parameters .	mm .	Points .
Diameter of Ao ascendens	<45	0
	45–54	1
	≥55	2
Length of Ao ascendens (3D-central line; AV to BCT)	<120	0
	≥120	1

Parameters .	mm .	Points .
Diameter of Ao ascendens	<45	0
	45–54	1
	≥55	2
Length of Ao ascendens (3D-central line; AV to BCT)	<120	0
	≥120	1

AV: aortic valve annulus; BCT: brachiocephalic trunk; TAIPAN: Tübingen Aortic Pathoanatomy.

Prophylactic ascending aorta replacement at ≥ 2 points.

Table 3:

Open in new tab

Parameters of the TAIPAN score

Parameters .	mm .	Points .
Diameter of Ao ascendens	<45	0
	45–54	1
	≥55	2
Length of Ao ascendens (3D-central line; AV to BCT)	<120	0
	≥120	1

Parameters .	mm .	Points .
Diameter of Ao ascendens	<45	0
	45–54	1
	≥55	2
Length of Ao ascendens (3D-central line; AV to BCT)	<120	0
	≥120	1

AV: aortic valve annulus; BCT: brachiocephalic trunk; TAIPAN: Tübingen Aortic Pathoanatomy.

Prophylactic ascending aorta replacement at ≥ 2 points.

CONCLUSION

The ascending aorta and the aortic arch in dissected and pre-dissection aortas are significantly elongated compared to healthy controls. Aortic elongation may be measured using the curved multiplanar reformation method and a score composed of ascending aortic dilatation and elongation may facilitate a better risk stratification for TAD. The pathophysiological and clinical relevance of aortic elongation will be subject of further investigation. In future studies, the number of preTAD patients must be increased to allow for a more sophisticated analysis of the aortic risk morphology and to evaluate the clinical value of the proposed score.

Funding

This work was supported by the Dr Karl Kuhn-Stiftung, Tübingen, Germany.

Conflict of interest: none declared.

REFERENCES

Author notes