Benjamin Oliver Patterson, BS, MRCS, Alberto Vidal-Diez, BS, Alan Karthikesalingam, PhD, MRCS, Peter J. E. Holt, PhD, FRCS, Ian M. Loftus, MD, MRCS, and Matt M. Thompson, MD, MRCS St. George’s Vascular Institute, St. George’s Hospital, London, United Kingdom
Background. Different methods have been used to assess remodeling of the thoracic aorta after endovascular treatment of Stanford type B aortic dissections. Changes in morphology may be described using diameter, area, or volume. The aim of this study was to determine if aortic diameter measurements could be used to approximate aortic area in order to refine reporting standards. Methods. The study population encompassed 100 patients enrolled in the VIRTUE registry (designed to assess thoracic endografting with the Valiant Stent Graft System [Medtronic, Minneapolis, MN] for the treatment of type B aortic dissections). Diameter and area measurements of the true lumen, false lumen, and whole aorta were made using three-dimensional computed tomographic (3D CT) workstations, at different anatomic locations. Measurements included preoperative, postoperative, and followup scans. The Pearson test was used to determine general correlation between diameter and volume at each location. Scatter plots were drawn and linear regression
models were used to draw a line of best fit. Comparison of these with nonlinear models was performed. Results. Aortic true and false lumen diameter and area showed good correlation (p < 0.001) in the majority of anatomic locations. This relationship was present preoperatively and during follow-up (p < 0.001). The linear regression models fit well with high R2 values. At very large aortic sizes nonlinear models were a slightly better fit, but this was not significant. Conclusions. Aortic diameter measurements correlate with luminal areas in patients with type B aortic dissection. This implies area increases proportionately with diameter over time. Therefore, diameter measurements using multiplanar reconstructions based on a central luminal line appear to be adequate when assessing aortic remodeling after endovascular treatment of aortic dissection.
V
these measurements are time consuming and are not part of routine clinical practice. Area and volume estimation are also subject to greater interobserver variability, whereas aortic diameters can be measured with excellent reproducibility [6]. The aim of this study was to determine the relationship between area and diameter in both the true and false lumen. A linear relationship between luminal diameter and area would suggest that area increases in proportion to diameter, and allow the formulation of pragmatic morphology reporting standards based on aortic diameters.
arious measurements have been used to describe morphologic changes in the thoracic aorta after endovascular treatment of Stanford type B dissection [1]. These vary from simple diameter measurements from axial scans to complex volumetric studies of the true and false lumen [2, 3]. Recent evidence has emerged to suggest that favorable aortic remodeling after endovascular treatment is associated with a positive clinical outcome [4, 5]. A significant limitation of existing studies is that most report different parameters over variable follow-up intervals [1]. There are no sufficiently large studies to allow analysis of factors such as length of coverage, timing of treatment, and preoperative morphology that may potentially influence remodeling. A standardized protocol that could be reported for all studies of aortic morphology would allow for more detailed analysis and ideally would contain readily obtainable measurements. Measuring luminal areas and volumes would seem to be the most sensitive way to detect changes in aortic morphology, but
(Ann Thorac Surg 2015;99:95–102) Ó 2015 by The Society of Thoracic Surgeons
Material and Methods Patient Population and Morphology Data The study population consisted of patients enrolled into the VIRTUE registry, which is described in detail elsewhere [7]. The VIRTUE registry consists of a total of 100 patients who were treated with the Valiant endovascular stent-graft
Accepted for publication Aug 6, 2014. Address correspondence to Dr Patterson, St. George’s Vascular Institute, 4th Flr, St. James Wing, St. George’s Hospital NHS Trust, London SW17 0QT, United Kingdom; e-mail:
[email protected].
Ó 2015 by The Society of Thoracic Surgeons Published by Elsevier
Dr Thompson discloses a financial relationship with Medtronic.
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2014.08.022
ADULT CARDIAC
Comparison of Aortic Diameter and Area After Endovascular Treatment of Aortic Dissection
ADULT CARDIAC
96
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
system (Medtronic, Minneapolis, MN) for aortic dissection. Fifty patients were treated for acute aortic dissection within 2 weeks of presentation; 24 for subacute dissection 15 to 92 days, and 26 treated for chronic dissection more than 92 days after the initial diagnosis. Patients were followed up at intervals of 6, 12, 24, and 36 months. Informed consent was gained from all patients enrolled and the study was approved by local ethical review boards in all cases (clinicaltrials.gov identifier number NCT01213589). Computed tomographic scans obtained at each follow-up visit were measured by a core lab according to an agreed protocol using the 3Mensio system (3Mensio Medical Imaging BV, Bilthoven, The Netherlands). This system has previously been validated for use when measuring morphology in the infrarenal aorta with excellent intra and interobserver agreement [6]. The area and diameters of the whole aorta (WA), true lumen (TL), and false lumen (FL) were measured at the ascending aorta, the left subclavian artery (LSA), 20-mm beyond the LSA, 100-mm beyond the LSA, the celiac axis, and at the point of maximum aortic diameter within the descending thoracic aorta (Fig 1). A central luminal line was installed within the aorta and “stretch views” were obtained. Multiplanar reconstructions of these images in the plane orthogonal to the central luminal line allowed for accurate estimates of diameter and area by correcting for parallax error (Fig 2A, B). The area of the WA was measured using the manual area calculation tool
Ann Thorac Surg 2015;99:95–102
to draw around the outside of the aortic wall as it appeared on the multiplanar reconstructions image. The area of the TL and FL were measured by drawing around the corresponding lumens. The diameter of the WA was measured from outside wall to outside wall through the central luminal line and taking the greatest measurement obtained. The diameters of the TL and FL were taken using the same line but only measuring the relevant lumen.
Statistical Analysis Statistical analysis was performed using the SAS 9.3 statistical package (SAS Institute Inc, Cary, NC) and SPSS 20 (IBM, Armonk, NY). Graphs were drawn with R 3.0.1. Mean and median diameters and standard deviations were calculated for each anatomic point for the WA, TL, and FL. The Pearson correlation coefficient was used to compare the area and diameter measurements for the WA, the FL, and the TL in the different anatomic locations at the different follow-up intervals with significance testing to the p less than 0.05 level of stringency. The purpose of this analysis was to determine if there was a correlation between the 2 measurements in each anatomic location and to see if this relationship was consistent over follow-up. Observations with zero value for area and diameter were removed from the calculation to avoid inflating the correlation. Scatter plots were produced to determine the nature of the relationship between diameter and area. Linear regression models were fitted using only the corresponding measures of diameter as a fixed effect. The within patient correlation was estimated using scan time as random effect with an unstructured variance and covariance matrix. The R-square value, which refers to the fraction of variance explained by the model, of each model was reported. Finally, a comparison was made with both exponential and quadratic regression models to see if these accounted for a potentially nonlinear or curved distribution, and more accurately reflected the distribution of the scatter plots.
Results
Fig 1. Location of area and diameter measurements of the thoracic aorta: 1 ¼ maximum ascending aorta diameter; 2 ¼ at the left subclavian artery; 3 ¼ at 2 cm beyond the left subclavian artery; 4 ¼ 10 cm beyond the left subclavian artery, 5 ¼ the widest part of the descending thoracic aorta; and 6 ¼ at the level of the celiac axis.
There were 82 patients who had analysis of preoperative scans and 81 with discharge scans available for analysis. At follow-up there were 71, 66, 64, and 61 scans at 6, 12, 24, and 36 months, respectively. There were 26 patients who had scans performed at other time intervals for unscheduled visits. The Pearson test for correlation between area and diameter was performed for all measurements at all time periods between diameter and area measurements taken at the same anatomic location, as was the R2 value obtained from the linear regression model. A sample of 14 scans were tested for interobserver variability in area, with a mean and median of 2% in the relative difference between 2 observers and a standard deviation of 9% noted. There was 1 episode of a significant disagreement between measurements and this was discussed and resolved by consensus at an investigators meeting.
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
97
Fig 2. (A) Views used to measure the thoracic aorta using the 3-mensio system. The central luminal line is installed which (left) allows a “stretch view” to be produced (left). This allows accurate determination of anatomic landmarks and precise location of measurements. (B) Diameters and areas are measured using multiplanar reconstructions of the computed tomographic images in the plane perpendicular to the central luminal line using the appropriate measurement tools.
Whole Aorta Measurements
False Lumen Measurements
There was excellent correlation between whole aortic area and diameter measurements at all time points, at all anatomic locations for all scheduled scans (p < 0.001) (Table 1). The linear regression model generated a line of best fit that suggested that WA diameter was increased in proportion to area until the aorta was over approximately 90 mm in diameter. At this point, area appeared to increase more rapidly than would have been expected (Fig 3A). The R2 values for the linear regression were high, with a mean of 86.7% in the ascending aorta. The lowest mean R2 value was 70.1% and was at the level of the ostium of the LSA. The plots of observed values of area observed versus those predicted by the model using diameter as a coefficient showed close agreement (Fig 3B).
In the ascending aorta and at the level of the LSA there were not enough patients with a false lumen described to allow meaningful analysis. In locations distal to the LSA, correlation was excellent between diameter and area (p < 0.005) (Table 2). The scatter plots drawn for the FL showed a linear relationship between diameter and area, and this was reflected in the line of best fit that was derived (Fig 4A). Once again there was some loss of this linear relationship when the false lumen was over 50 mm. The mean R2 values were generally good but lowest at the level of the celiac axis (56.1%). The plots of observed values of area versus those predicted by the model using diameter as a coefficient showed close agreement (Fig 4B).
ADULT CARDIAC
Ann Thorac Surg 2015;99:95–102
ADULT CARDIAC
98
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
Ann Thorac Surg 2015;99:95–102
Table 1. Correlation Between Area and Diameter Measurements for the Whole Aortaa Variable Ascending aorta
LSA
LSA D 2 cm
LSA D 10 cm
Celiac axis
Descending thoracic aorta
a
Scan Interval /Months
n
Pearson’s Correlation
p Value
R2
Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36
82 81 71 66 64 61 81 78 72 65 60 61 78 78 68 65 63 57 81 81 71 65 64 59 76 77 70 64 61 61 78 78 68 62 61 62
79.35% 98.57% 96.40% 96.70% 92.10% 98.54% 90.23% 73.75% 48.35% 86.55% 95.44% 96.77% 85.01% 88.81% 72.72% 93.35% 95.39% 96.93% 78.62% 83.79% 93.16% 97.04% 95.93% 97.14% 87.09% 89.87% 82.71% 91.02% 95.71% 96.83% 95.37% 88.04% 96.06% 82.51% 97.22% 96.65%
<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
62.96% 97.16% 92.93% 93.50% 84.82% 97.09% 81.41% 54.38% 23.37% 74.90% 91.08% 93.64% 72.26% 78.87% 52.88% 87.15% 90.98% 93.96% 61.80% 70.21% 86.78% 94.16% 92.03% 94.37% 75.85% 80.77% 68.42% 82.85% 91.61% 93.75% 90.95% 77.52% 92.27% 68.07% 94.51% 93.41%
p < 0.0001 in all cases denoting excellent correlation; the R2 values for the subsequent regression models are also displayed.
LSA ¼ left subclavian artery;
Postop ¼ postoperative;
Preop ¼ preoperative.
True Lumen Measurements There were many patients in whom there was no FL proximal to the LSA on the preoperative scan and therefore no corresponding measure of TL was made, so these were not included in the analysis. Distally to this the correlation between diameter and area was excellent for all measurements (p < 0.001 in all but 1 case) (Table 3). The scatter plots drawn for the TL showed a linear relationship between diameter and area and this was reflected in the line of best fit that was derived (Fig 5A). Visually this linear relationship was not as clear at sizes over approximately 37 mm. The mean R2 value was generally good but lowest at the celiac axis (59.1%). The plots of observed values of area versus those predicted by the model using diameter as a coefficient showed close agreement (Fig 5B).
Comparison of Linear, Quadratic, and Exponential Models Linear, quadratic, and exponential regression models were used to derive lines of best fit for the scatter plots of diameter versus area combining all follow-up time points (Table 4). The linear model displayed an acceptable R2 value in all locations. The quadratic curve appeared to be the best fit overall due mainly to the larger aortas that tended to have areas greater than would be predicted by the regression models. This was especially true for the FL measurements. Even in locations such as the celiac axis, where the linear relationship between diameter and area was seen to be less pronounced according to the R2 statistic, the curves were very similar (Fig 6).
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
99 ADULT CARDIAC
Ann Thorac Surg 2015;99:95–102
Fig 3. (A) Scatter plot of area and diameter in the descending thoracic aorta (DTA) plotted with the line of best fit according to the linear model displayed. (B) A scatter plot of observed diameter (black) values plotted against the fitted values of the regression model from the descending thoracic aorta based on corresponding diameter.
Comment This study demonstrates that aortic diameter measurements taken perpendicular to an installed central luminal line are related to area in a linear fashion over multiple
time points. This implies that changes in diameter measurements accurately reflect changes in area, which would likely be considered a better proxy measurement of blood flow through a lumen. Diameter measurements at key anatomic locations could therefore form the basis
Table 2. Correlation Between Area and Diameter Measurements for the False Lumena Variable LSA D 2 cm
LSA D 10 cm
Descending thoracic aorta
Celiac axis
a
Scan Interval / Months
n
Pearson Correlation
p Value
R2
Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36
55 23 60 19 15 9 75 76 44 36 27 24 75 74 46 38 39 34 62 61 44 38 27 25
71.92% 86.38% 86.09% 89.71% 88.23% 96.05% 70.60% 87.52% 78.56% 93.34% 91.08% 94.52% 82.00% 84.18% 91.98% 88.98% 80.85% 93.73% 66.13% 87.98% 56.55% 75.14% 77.85% 87.13%
<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
51.72% 74.61% 74.11% 80.48% 77.84% 92.26% 49.84% 76.59% 61.72% 87.13% 82.96% 89.34% 67.24% 70.87% 84.60% 79.17% 65.37% 87.86% 43.73% 77.40% 31.98% 56.46% 60.61% 75.91%
p < 0.0001 in all cases denoting excellent correlation; the R2 values for the subsequent regression models are also displayed.
LSA ¼ left subclavian artery;
Postop ¼ postoperative;
Preop ¼ preoperative.
ADULT CARDIAC
100
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
Ann Thorac Surg 2015;99:95–102
Fig 4. (A) Scatter plot of area and diameter 10 cm from the left subclavian artery (LSA) plotted with the line of best fit according to the linear model displayed. (B) A scatter plot of observed diameter (black) values plotted against the predicted values of the regression model 10 cm from the LSA based on corresponding diameter. (FL ¼ false lumen.)
of a pragmatic minimal required dataset to be used as a guide when designing protocols, and reporting registries or institutional series. Establishing such a protocol is important given recent evidence suggesting that aortic remodeling is key in preventing aortic-related complications in patients with dissection [5]. The advantage of
diameter measurements is that they can be performed quickly in a highly reproducible fashion using all commercially available 3D CT reconstruction software. This has been demonstrated directly in the derivation of a protocol for assessing the morphology of the infrarenal aorta [6].
Table 3. Correlation Between Area and Diameter Measurements for the True Lumena Variable LSA D 2 cm
LSA D 10 cm
Descending thoracic aorta
Celiac axis
a
Scan Interval / Months
N
Pearson Correlation
p Value
R2
Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36 Preop Postop 6 12 24 36
56 64 26 19 44 59 75 78 45 37 49 62 75 75 48 37 54 62 63 62 47 38 50 61
39.29% 74.92% 70.26% 90.62% 90.04% 83.59% 81.79% 80.99% 77.61% 57.81% 72.75% 91.74% 83.97% 70.77% 93.77% 92.22% 64.56% 92.28% 60.11% 85.83% 54.94% 70.24% 85.42% 92.29%
0.0027 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
15.44% 56.13% 49.36% 82.11% 81.07% 69.87% 66.90% 65.60% 60.23% 33.42% 52.93% 84.15% 70.52% 50.08% 87.93% 85.04% 41.68% 85.16% 36.13% 73.67% 30.18% 49.33% 72.97% 85.18%
p < 0.0001 in all cases denoting excellent correlation; the R2 values for the subsequent regression models are also displayed.
LSA ¼ left subclavian artery;
Postop ¼ postoperative;
Preop ¼ preoperative.
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
101
Fig 5. (A) Scatter plot of the true lumen (TL) area and diameter at the celiac trunk (CT) plotted with the line of best fit according to the linear model displayed. (B) A scatter plot of observed TL diameter (black) values plotted against the predicted values of the regression model at the celiac axis based on corresponding diameter.
An extremely important caveat is that these conclusions are only valid when using measurements that are derived using three-dimensional reconstruction of CT images by installing a central luminal line and creating multiplanar reconstructions. If this is not done the influence of “parallax error” means that diameter measurements cannot be relied on. Recent work has confirmed that measuring diameters using axial images alone is inadequate for clinical assessment and therefore should not be performed as part of data gathering for research purposes [8]. The present study found that the linear relationship between diameter and area appear to become less accurate in large aortas, specifically in the false lumen. At false
lumen, diameters greater than approximately 40 mm the corresponding area tended to be larger than would be predicted by a linear relationship with diameter (Fig 3A, B). This may be because the false lumen can adopt a variety of different eccentric shapes at larger sizes. The implication of this is that diameter measurements alone should not be used when making decisions regarding the management of individual patients in patients with complex false lumen configurations. Subgroup analyses revealed that the linear relationship between area and diameter was present regardless of the chronicity of the presenting pathology, but further analysis of the difference in the rate of aortic remodeling was beyond the scope of this investigation.
Table 4. A Comparison of Linear, Quadratic and Exponential Models Fitted to the Diameter vs Area Scatter Plots Including All Time Periodsa R2 Location and Measurement Ascending aorta WA LSA ostium WA LSA 20 mm
LSA 100 mm
Descending thoracic aorta
Celiac trunk
a
WA WA WA FL TL WA FL TL WA FL TL WA FL TL
Linear
Quadratic
Exponential
0.887 0.631 0.746 0.72 0.546 0.825 0.691 0.702 0.856 0.754 0.683 0.804 0.7 0.656
0.887 0.652 0.747 0.722 0.562 0.851 0.746 0.742 0.88 0.79 0.722 0.81 0.717 0.718
0.823 0.599 0.742 0.718 0.544 0.788 0.721 0.703 0.752 0.732 0.645 0.798 0.672 0.627
The coefficients for all models were significant to the p < 0.001 level.
FL ¼ false lumen;
LSA ¼ left subclavian artery;
TL ¼ total lumen;
WA ¼ whole aorta.
ADULT CARDIAC
Ann Thorac Surg 2015;99:95–102
ADULT CARDIAC
102
PATTERSON ET AL ASSESSING REMODELING AFTER TREATED DISSECTION
Ann Thorac Surg 2015;99:95–102
area. This knowledge will help in the development of a pragmatic minimum required dataset that can be used to report on studies of aortic dissection. We would like to acknowledge Debra Shaver and Victoria Rendon of Medtronic for providing the registry data and assisting with the study. The views expressed in this publication are those of the authors and not necessarily those of the National Health Service, the National Institute for Health Research, or the UK Department of Health.
References
Fig 6. A comparison of different regression models fitted to a scatter plot of diameter versus area. Coefficients for each model were significant to the p < 0.001 level. (B ¼ observed; — ¼ linear; — ¼ quadratic; – ¼ exponential.)
A limitation of this study is that the VIRTUE registry featured only patients who were treated with Valiant endovascular stent grafts, and therefore cannot comment on the follow-up of patients treated with medical therapy alone. Despite this, preoperative scans were available and the linear relationship between diameter and area was present in this group. Comparing diameter and area to volume at different anatomic locations would also have been useful, but these data were not recorded in the original registry. In patients with Stanford type B dissection, the diameters of the true lumen, false lumen, and whole aorta accurately reflect the corresponding areas at key anatomic location. The linear relationship observed means that increasing diameter over time corresponds to increasing
1. Patterson BO, Cobb RJ, Karthikesalingam A, et al. A systematic review of aortic remodeling after endovascular repair of type B aortic dissection: methods and outcomes. Ann Thorac Surg 2014;97:588–95. 2. Kusagawa H, Shimono T, Ishida M. Changes in false lumen after transluminal stent-graft placement in aortic dissections: six years’ experience. Circulation 2005;111:2951–7. 3. Qing KX, Yiu WK, Cheng SW. A morphologic study of chronic type B aortic dissections and aneurysms after thoracic endovascular stent grafting. J Vasc Surg 2012:551268–76. 4. Kim KM, Donayre CE, Reynolds TS, et al. Aortic remodeling, volumetric analysis, and clinical outcomes of endoluminal exclusion of acute complicated type B thoracic aortic dissections. J Vasc Surg 2011;54:316–25. 5. Nienaber CA, Kische S, Rousseau H, et al. Endovascular repair of type B aortic dissection: long-term results of the randomized investigation of stent grafts in aortic dissection trial. Circ Cardiovasc Interv 2013;6:407–16. 6. Ghatwary T, Karthikesalingam A, Patterson B, et al. St. George’s Vascular Institute Protocol: an accurate and reproducible methodology to enable comprehensive characterization of infrarenal abdominal aortic aneurysm morphology in clinical and research applications. J Endovasc Ther 2012;19: 400–14. 7. Investigators TVR. The VIRTUE Registry of type B thoracic dissections–study design and early results. Eur J Vasc Endovasc Surg 2011;41:159–66. 8. Rudarakanchana N, Bicknell CD, Cheshire NJ, et al. Variation in maximum diameter measurements of descending thoracic aortic aneurysms using unformatted planes versus images corrected to aortic centerline. Eur J Vasc Endovasc Surg 2014;47:19–26.