Author + information
- Received April 2, 2012
- Revision received June 13, 2012
- Accepted July 26, 2012
- Published online March 1, 2013.
- Aditya S. Shirali, BS⁎,⁎ (, )
- Moritz S. Bischoff, MD⁎,
- Hung-Mo Lin, PhD⁎,
- Irina Oyfe, MS†,
- Robert Lookstein, MD†,
- Randall B. Griepp, MD⁎ and
- Gabriele Di Luozzo, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Aditya S. Shirali, Cardiothoracic Surgery, Icahn School of Medicine at Mount Sinai, 1190 Fifth Avenue, New York, New York 10029
Objectives This study sought to identify possible anatomic predictors of acute type B aortic dissection (AAD) in hypertensive patients using multidetector computed tomography angiography (CTA).
Background Although hypertension remains one of the most significant risk factors for AAD development, it is unlikely to be the only risk factor for AAD. Few studies have assessed anatomical predictors of AAD development.
Methods CTA of normotensive patients without AAD (group 1, n = 35), hypertensive patients without AAD (group 2, n = 37), and hypertensive patients with AAD (group 3, n = 37) were compared. The length, diameter, volume, and tortuosity of the aorta as well as arch vessel angulation were measured for each patient and normalized to group 1 averages. Stepwise logistic regression identified significant anatomical associations; the model was validated based on 1,000 bootstrapped samples.
Results The demographics of the groups were similar. The length of the proximal and entire aorta, the diameters in the proximal ascending aorta and aortic arch, and the aortic volumes were all greater (p < 0.0001, p = 0.0064 for ascending aortic diameter) in group 3 than in groups 1 and 2, as was entire aortic tortuosity (p < 0.0001). An AAD risk model was developed based on aortic arch diameter, length from the aortic root to the iliac bifurcation, and angulation of the brachiocephalic artery origin from the aorta. The bootstrap estimate of the area under the receiver operating curve was 0.974.
Conclusions Enlargement of the ascending aorta and aortic arch and increased aortic tortuosity reflect an aortopathy which enhances the probability of AAD. A model based on 3 anatomical variables demonstrates significant associations with AAD: it may allow identification by aortic imaging of the hypertensive patient most at risk, and permit implementation of aggressive medical management and consideration of pre-emptive surgery to prevent dissection.
Acute type B aortic dissection (AAD) is a highly feared thoracic aortic pathology encountered in hospital settings. Although many type B AADs can be managed medically, most patients with this pathology suffer significant morbidity and mortality (1). Even after optimal medical and surgical therapy, the long-term prognosis of type B AAD is dismal, with 5- and 10-year survival rates of 60% and 35%, respectivley (2,3).
Hypertension remains one of the most significant risk factors for type B AAD development, with nearly 75% of patients having a history of hypertension (4). Nevertheless, although nearly 75 million adults in the United States are hypertensive, the incidence of AAD is only 2.9 to 3.5 per 100,000 person-years (5,6). Given the high mortality rate and dismal long-term prognosis of type B AAD, it seems logical to try to identify additional characteristics that might predispose an individual with hypertension to development of type B AAD. This study is an attempt to provide an anatomic aortic profile to identify those hypertensive patients most at risk, who may need more aggressive medical management or even pre-emptive surgery to prevent dissection.
A review of our institutional database disclosed 56 patients with AAD who presented to Mount Sinai Medical Center (MSMC) from February 2002 to July 2010. A multidetector computed tomography angiography (CTA) within 2 weeks of presentation and symptom onset was required to diagnose AAD: patients with type B AAD were excluded from the study if CTA was not carried out as part of the diagnostic examination, if the study was done at an outside institution, if the dissection was secondary to a traumatic event, or evidence of congenital or connective tissue disease was found, such as bicuspid aortic valve or Marfan's syndrome. These criteria resulted in inclusion of 37 type B AAD patients. Data from these patients were compared with those for 37 hypertensive patients without AAD and those for 35 normotensive patients without AAD who underwent CTA.
The institutional review board approved this retrospective research and waived the need for individual patient consent. Patient information was obtained from medical records of patients presenting to the MSMC Emergency Department (ED) with AAD or who were transferred to our institution after confirmation of diagnosis. Normotensive and hypertensive adult patients who presented to the ED with episodes of chest, abdominal, or back pain described as sharp, pressure-like, or aching and who underwent CTA to determine the cause of the presenting symptoms were included in the study. ED medical records of these patients were then assessed for patients' presenting blood pressure and antihypertensive medication history.
Electrocardiographic (ECG)-gated CTA of the thoracic aorta was carried out with injection of 100 ml of contrast with 50 ml of saline chaser, threshold 80 HU; rotation speed: 330 ms; collimation: 64 × 0.6 mm; pitch: 0.2; voltage: 140 kV; current: 700 to 850 mAs. Datasets were analyzed on a cardiovascular workstation (AquariusWS version 188.8.131.52, TeraRecon, San Mateo, California) using dedicated vascular analysis software. The software enables real-time diagnostic review of 2-, 3-, and 4-dimensional images for managing large thin-slice computed tomography and magnetic resonance scans, and includes workflow tools which simplify the interpretation by automatically presenting the 3-dimensional volume based on the study type selected by the operator.
The length of the ascending aorta from the aortic root to the left subclavian artery, and the entire aorta from the aortic root to the iliac bifurcation were measured (Fig. 1). Additional measurements included the volumes of the ascending aorta and the aortic arch, as well as the maximal diameters of the aortic root, the ascending aorta at the right pulmonary artery, and the aortic arch. Tortuosity of the aorta, defined as the length of the midline within the aorta divided by the linear distance between the aortic root and the iliac bifurcation, was also calculated. The tortuosity of the ascending aorta, defined as the measured length of the ascending aorta divided by the linear distance between the aortic root and left subclavian artery, was similarly calculated for each patient (Figs. 2A and 2B). The angle of the origin of the brachiocephalic and of the left subclavian arteries from the aortic arch was measured. The volume and diameter measurements of the descending and abdominal aorta were likely confounded by the presence of type B AAD, and were not compared between the groups. Analysis was therefore restricted to the ascending aorta and aortic arch proximal to the left subclavian artery.
For group comparisons, we used chi-square tests or Fisher exact tests for categorical variables, analysis of variance (ANOVA) allowing for heterogeneous group variances using PROC MIXED in SAS version 9.2 software (SAS Institute Inc., Cary, North Carolina), or nonparametric Kruskal-Wallis tests for continuous variables. For pairwise comparisons in Tables 1 and 2, we were interested in comparing the hypertensive group against normotensive and AAD groups. We have reported the individual pairwise p values. One may exercise judgment to correct for multiple comparisons by multiplying the p values by the number of comparisons (e.g., 2). For most pairwise comparisons, the conclusions are the same with or without adjustment of multiple comparisons. The correlation between ascending aortic volume and diameter and systolic blood pressure was evaluated by Pearson correlation coefficient. Stepwise logistic regression analyses were performed to identify significant predictors of AAD. The covariates considered in the initial model included measurements of diameter, length, tortuosity, and angulation listed in Table 3. Results were reported as odds ratios and 95% confidence intervals for the identified risk factors.
We used the bootstrap method to validate the model selection (7). For each sample the same stepwise model selection procedure was used with entry and stay criteria set to be 0.1, leading to potentially different selection of significant AAD anatomic variables. The procedure robustness was evaluated by the consistency of the selected covariates. The bootstrap estimate of the coefficient for AAD was then compared to that from the model using all the data. The model's strength selected by this procedure was assessed by the area under the receiver-operating curve (ROC) by using the nonparametric “0.632” estimate value. This estimate uses a weighting of 0.368 × apparent + 0.632 × average (test) to correct for overoptimism in the sample estimate, where apparent is the estimate from the entire dataset, and test is the estimate from patients not selected in the bootstrap sample (8,9). All statistical analyses were carried out using SAS version 9.2 (SAS Institute, Inc.).
The patient groups showed no difference in age, sex, or body mass index (Table 1). Table 3 summarizes the normotensive patient anatomical data. Compared with normotensive patients, the lengths of the proximal and entire aorta were 101% and 101% in hypertensive patients without AAD and 117% and 120% in patients with AAD (Table 2). The proximal and entire aortic lengths were significantly larger in AAD patients compared with both normotensive and hypertensive patients (p < 0.0001), but statistically insignificant between hypertensive and normotensive patients without AAD.
Volumetric analysis showed significant differences among the 3 groups in the ascending aorta and aortic arch (p < 0.0001). The volumes of the ascending aorta and aortic arch in hypertensive patients were 123% and 110% those of normotensive patients. The volumes in AAD patients were 175% and 167% those of normotensive patients (Table 2). Both the ascending aortic and aortic arch volumes were significantly larger in AAD patients than in either normotensive or hypertensive patients (p < 0.0001), and hypertensive patients had significantly larger ascending aortic volumes than normotensive patients (p = 0.0068). The correlation coefficient for the relationship between ascending aortic volume and systolic blood pressure is 0.33 (p = 0.0005), suggesting a positive relationship between systolic blood pressure and ascending aortic volume.
Diameter analysis similarly showed significant differences among the 3 groups at the aortic root, ascending aorta (at right pulmonary artery), and aortic arch (p < 0.0001). The aortic root, ascending aorta, and aortic arch diameters of hypertensive patients were 105%, 113%, and 104% those of normotensive patients. The maximal diameters of AAD patients at these 3 locations were 121%, 124%, and 139% compared to those of normotensive patients (Table 2). The aortic root, ascending aortic and aortic arch diameters were all significantly larger in AAD patients compared with both normotensive and hypertensive patients (p < 0.0001, p = 0.0064 for ascending aorta). The difference in aortic diameters between hypertensive and normotensive patients was statistically insignificant except at the ascending aorta (p < 0.0001). The correlation coefficient associated with the relationship between ascending aortic diameter and systolic blood pressure is 0.31 (p = 0.0012), suggesting a positive relationship between ascending aortic diameter and systolic blood pressure.
No significant differences between groups were observed in the tortuosity of the proximal aorta or in the left subclavian artery angle. There was, however, a significant difference between the 3 groups in tortuosity of the entire aorta (p < 0.0001): tortuosity in hypertensive patients was 102%, and in AAD patients 114% that of normotensive patients (Table 2). Tortuosity of the entire aorta was significantly greater in AAD patients than in normotensive and hypertensive patients (p < 0.0001), but statistically insignificant between hypertensive and normotensive patients. In addition, there was a significant difference between the three groups in the brachiocephalic artery angle (p = 0.0011) with angulation of 86% in hypertensive patients and 80% in AAD patients. Brachiocephalic artery angulation was significantly lower in hypertensive patients than in normotensive patients (p = 0.02), but insignificant between hypertensive and AAD patients.
Risk model using anatomic variables
A multivariable model was developed to provide a means by which one might find aortic anatomic variables with strong association to AAD occurrence. Stepwise logistic regression identified aortic arch diameter, entire aortic length, and the brachiocephalic artery angle as variables with strong association, scaled by the means of the normotensive group presented in Table 3 and then multiplied by 100.
Table 4 illustrates multivariable models and model validation. The probability (Pr) of AAD is Pr(AAD) = , in which e refers to the natural exponential function and y is the linear combination of the values of the variables (relative to the normotensive group and multiplied by 100) multiplied by the β coefficients listed in Table 4. The equation in Model 1 based on the stepwise selection has strong predictability (area under the ROC). The increment in predictive power is from 0.909 (if only the diameter was used as the predictor) to 0.962 if length was also added, and 0.974 if brachiocephalic angle was included last. Model selection using stepwise logistic regression was repeated for 1,000 bootstrapped samples, leading to a potentially different selection of significant AAD anatomic variables. Aortic arch diameter appeared in at least 99% of the repetitions, followed by aortic length (81%), aortic tortuosity (56%), and brachiocephalic angle (48%). No other anatomic variables listed in Table 3 appeared more than 40% of the time. There was a slight improvement in ROC if the model included aortic tortuosity as an additional variable (area under ROC: 0.977). Therefore, validation using bootstrap methodology showed a high degree of consistency among the covariates included in Model 1.
Blood pressure and antihypertensive medication
Table 1 summarizes the systolic and diastolic blood pressures in the three patient groups. As expected, the groups had different systolic and diastolic blood pressures (p < 0.0001 and p = 0.01, respectively). Systolic pressures were significantly higher in AAD patients compared with both normotensive and hypertensive patients (p = 0.02) but the extent to which they were higher than in the hypertensive group was less than might have been anticipated perhaps because some of the patients had antihypertensive treatment, usually intravenous labetalol or clonidine, prior to admission to our institution.
Table 5 lists the differences in hypertension management between patients with AAD and hypertensive patients without AAD. Normotensive patients by definition were not on antihypertensive medications. Among AAD patients, 32 of 37 (87%) were taking antihypertensive medications, while 26 of 37 (70%) hypertensive patients without AAD were taking antihypertensive medications: this difference was insignificant. Despite the somewhat higher blood pressures recorded in AAD patients, no difference was found in the number of antihypertensive medications being taken between the 2 groups.
The present study devised an aortic profile based on retrospective CTA analysis that paves the pathway for predicting AAD occurrence. CTA imaging provides excellent diagnostic accuracy for AAD (between 88% and 100%) and allows the full extent of the acute dissection to be visualized noninvasively (10). Using the IRAD (International Registry of Aortic Dissection) database, Moore et al. (11) found that CTA has a sensitivity of 93% for diagnosing type B AAD: it is currently the most frequently used imaging modality for diagnosing AAD and most often the initial test to be done when AAD is suspected.
Few studies have explored the use of CTA to provide measurements that may be used for prevention of dissection, however. Mathematical models to predict type A AAD development have been constructed (12), but few studies have found conclusive anatomic evidence to predict type B AAD. Using the IRAD database, Pape et al. (13) found that ascending aortic diameter >5.5 cm could not adequately predict risk of type A AAD. The database was further used to retrospectively assess type B AAD risk, which found that descending aortic diameter >5.5 cm could not predict type B AAD, with over 80% of patients with diameters <5.5 cm developing dissection (14). Both studies concluded that aortic diameter measurements are not useful parameters for prevention of AAD, yet both retrospectively measured aortic segments within the dissected area, decreasing the utility and reliability of the measurements. The current study, however, uses anatomic variables proximal to the dissection area to associate with type B AAD risk and few measurements of aortic segments distal to the left subclavian artery were used because they are likely distorted by the presence of dissection and are therefore useless when considering prediction. In considering the aorta proximal to the left subclavian artery, we found that both volume and diameter of the ascending aorta are increased in hypertensive patients without AAD, and further increased in AAD patients, who also present with significantly higher systolic blood pressures. These findings conform to current literature, which demonstrates that ascending aortic enlargement is common among patients presenting with type B AAD. In a large retrospective study, Booher et al. (15) find that over 40% of type B AAD patients had a proximal aorta that measured >4.0 cm and suggest that the proximal aortic dilatation may be a marker of the diffuse nature of aortic disease. We go on to find a positive correlation between the volume and diameter of the ascending aorta and systolic blood pressure, suggesting that the proximal aorta may not only serve as a potential marker for aortic disease, but also as one for poorly controlled hypertension. Although the use of ascending aortic diameter as a marker for poorly controlled hypertension is tempting due to the ease with which it may be measured via echocardiography, further investigation with a larger sample size is warranted before specific recommendations can be given.
The lack of association between aortic diameter at the dissection site and AAD development suggest that the search for measures of AAD risk must go beyond conventional measures of aortic diameter. We find that the increased length of the ascending aorta in AAD patients compared with normotensive and hypertensive patients, and the lack of a difference in this length between hypertensive and normotensive patients, provides a strong association that warrants prospective testing to determine its utility as a predictor of AAD risk. Increased length of the ascending aorta in the setting of hypertension may induce disturbed or turbulent blood flow. Studies using computed tomography imaging to measure blood flow in patients with type B AAD have shown areas of highly disturbed and turbulent flow with high values of wall sheer stress around the area of the true lumen (16). In the setting of an already dilated proximal aorta, the increased length may serve to further disturb blood flow in the hypertensive patient and increase the chance of dissection.
We further provide a multivariable model using aortic arch diameter, entire aortic length, and brachiocephalic artery angle to assess their association with AAD development in hypertensive patients. Literature centered on type B AAD prediction has routinely cited the inability of one risk factor, namely aortic diameter, to predict dissection development. With this model, we provide the first attempt to combine several easily obtainable anatomic variables to assess AAD risk. This model has a predictability of 0.974 after internal validation, suggesting a strong association between these anatomic variables and AAD development. Of course, further use of the model in a prospective manner will be necessary to adequately assess the accuracy of this model's predictive power. Retrospectively using the model in patients with ascending aortic enlargement prior to the study date is limited by lack of widespread use of ECG-gated CTA for type B AAD. While the use of this current model may be tempting, the model was developed with retrospective data of hypertensive patients who developed AAD, and warrants prospective validation prior to use in a clinical setting. It may only be appropriate for hypertensive patients, rather than patients with connective tissue disorders, genetic mutations, or other medical conditions that develop AAD.
The central dogma with regard to AAD prevention has centered upon medical therapy to control hypertension and prevent AAD development. Analysis of our hypertensive and AAD patients, however, showed no difference in the number of patients in each group on antihypertensive medications and no difference in the number of medications being taken. Although these groups of patients had management of hypertension with the same number of antihypertensive medications, one group had higher systolic blood pressure and AAD. Whether or not these patients have an underlying aortopathy, more aggressive therapy may be required to prevent AAD in this specific subset of patients. In these cases, use of anatomic aortic profiles may identify those patients who require more aggressive medical therapy or prophylactic surgery.
In this study, we have shown that CTA analysis to assess length, volume, diameter, tortuosity, and brachiocephalic angulation of the proximal aorta provides a model that uses multiple anatomic variables to assess AAD risk. It would be tempting to recommend CTA of all hypertensive patients were it not for the significant burden of radiation and cost for the many with a low risk of AAD. A more feasible approach would be to stratify patients based on adequacy of hypertension control. For patients with adequate medical control, it may be feasible to monitor ascending aortic diameters via echocardiography, which is often part of the routine care involved in managing hypertension. Following the ascending aortic diameter, and altering hypertensive management based on diameter enlargement may be a more reasonable option for these patients. Those who show ascending aortic diameter enlargement via echocardiography despite traditional medical management may benefit from referral for CTA and calculation of AAD risk. Although assessment of ascending aortic diameter enlargement via echocardiography provides a safe alternative to CTA, its accuracy in diametric analysis and AAD dissection risk have yet to be determined and require further investigation in a prospective manner.
The current study makes associations based on data obtained in a small number of patients who had already developed AAD at the time of the encounter. The nature of the study led to exclusion of a specific subset of patients with morbidities that might have altered the geometry of this segment, leading to a further diminution of the already small number of patients being studied. In addition, use of ECG-gated CTA for analysis limited the sample size and study interval, as many patients with older studies or incompatible studies not amenable to analysis by the vascular software were excluded. Despite the bootstrap technique and the high predictability, there remains the possibility that the model is influenced by characteristics unique to the relatively small sample. Although the use of measurements proximal to the left subclavian and dissection increase the likelihood of obtaining accurate aortic dimensions, one cannot rule out retrograde changes in the proximal aorta during the 14-day interval deemed as acute. In addition, we cannot be sure whether or not the blood pressure reading we obtained reflects what may have been a transient hypertensive crisis, of unknown trigger, or is reasonably reflective of the usual measurements in a patient compliant with an ongoing blood pressure regimen. Although the sample size was limited by the above-mentioned factors and stepwise logistic regression is most routinely applied to larger samples, the findings are certainly hypothesis generating. Characterization of an anatomic aortic profile for hypertensive patients is informative and sets the stage for prospective investigation.
Type B AAD patients are more likely to have generalized enlargement of the proximal aorta compared to patients with hypertension alone. Using retrospective data, we have developed a multivariable model that uses strong associations between aortic anatomic variables obtained from CTA and AAD occurrence to provide a means by which to predict AAD, warranting prospective modeling. The presence in AAD patients of higher blood pressures than their hypertensive counterparts without AAD despite similar management of hypertension suggests the presence of an underlying aortopathy in AAD patients that require active monitoring and aggressive management, such as serial ascending aortic diameter measurement and multidrug antihypertensive regimens. Using anatomical variables, we have identified anatomic markers that may be used to identify patients who may be at increased risk for AAD.
This study was funded by the Ousman Akram Memorial Fund and the Mount Sinai Medical Center. Dr. Lookstein has consulted for Bayer Interventional and Cordis; and has received honoraria from Boston Scientific. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Pim J. de Feyter, MD, PhD, served as Guest Editor for this article.
- Abbreviations and Acronyms
- acute aortic dissection
- multidetector computed tomography angiography
- receiver-operating curve
- Received April 2, 2012.
- Revision received June 13, 2012.
- Accepted July 26, 2012.
- American College of Cardiology Foundation
- Lloyd-Jones D.,
- Adams R.J.,
- Brown T.M.,
- et al.
- Efron B.,
- Tibshirani R.J.
- Poullis M.P.,
- Warwick R.,
- Oo A.,
- Poole R.J.
- Pape L.A.,
- Tsai T.T.,
- Isselbacher E.M.,
- et al.