Author + information
- Received July 28, 2010
- Accepted August 3, 2010
- Published online October 1, 2010.
- Vikram Kurra, MD⁎,
- Michael L. Lieber, MS⁎,†,
- Srikanth Sola, MD∥,
- Vidyasagar Kalahasti, MD⁎,‡,
- Donald Hammer, MD‡,
- Stephen Gimple, MD¶,
- Scott D. Flamm, MD⁎,‡,
- Michael A. Bolen, MD⁎,‡,
- Sandra S. Halliburton, PhD⁎,‡,§,
- Tomislav Mihaljevic, MD‡,
- Milind Y. Desai, MD⁎,‡ and
- Paul Schoenhagen, MD⁎,‡,§,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Paul Schoenhagen, Cleveland Clinic Foundation, Cardiovascular Imaging, Imaging Institute, and Heart and Vascular Institute, 9500 Euclid Avenue, Desk J 1-4, Cleveland, Ohio 44106
Objectives We hypothesized that the extent of aortic atheroma of the entire thoracic aorta, determined by pre-operative multidetector-row computed tomographic angiography (MDCTA), is associated with long-term mortality following nonaortic cardiothoracic surgery.
Background In patients evaluated for cardiothoracic surgery, presence of severe aortic atheroma is associated with adverse short- and long-term post-operative outcome. However, the relationship between aortic plaque burden and mortality remains unknown.
Methods We reviewed clinical and imaging data from all patients who underwent electrocardiographic-gated contrast-enhanced MDCTA prior to coronary bypass or valvular heart surgery at our institution between 2002 and 2008. MDCTA studies were analyzed for thickness and circumferential extent of aortic atheroma in 5 segments of the thoracic aorta. A semiquantitative total plaque-burden score (TPBS) was calculated by assigning a score of 1 to 3 to plaque thickness and to circumferential plaque extent. When combined, this resulted in a score of 0 to 6 for each of the 5 segments and, hence, an overall score from 0 to 30. The primary end point was all-cause mortality during long-term follow-up.
Results A total of 862 patients (71% men, 67.8 years) were included and followed over a mean period of 25 ± 16 months. The mean TPBS was 8.6 (SD: ±6.0). The TPBS was a statistically significant predictor of mortality (p < 0.0001) while controlling for baseline demographics, cardiovascular risk factors, and type of surgery including reoperative status. The estimated hazard ratio for TPBS was 1.08 (95% confidence interval: 1.045 to 1.12). Other independent predictors of mortality were glomerular filtration rate (p = 0.015), type of surgery (p = 0.007), and peripheral artery disease (p = 0.03).
Conclusions Extent of thoracic aortic atheroma burden is independently associated with increased long-term mortality in patients following cardiothoracic surgery. Although our data do not provide definitive evidence, they suggest a relationship to the systemic atherosclerotic disease process and, therefore, have important implications for secondary prevention in post-operative rehabilitation programs.
In population-based studies of asymptomatic individuals and persons with suspected coronary artery disease, calcified aortic atheroma has been shown to be associated with increased risk of major cardiovascular events (1,2).
In patients evaluated for cardiothoracic surgery, calcified and noncalcified aortic atheroma has been assessed using intraoperative epiaortic ultrasound and transesophageal echocardiography. In these studies, the focal presence and severity of atheroma at the most diseased aortic segment has been established as a predictor of post-operative complications, as well of adverse long-term outcome (3–6). However, it remains unknown whether the extent and distribution of atheroma burden along the entire aorta is related to mortality.
Multidetector-row computed tomographic angiography (MDCTA) of the chest and aorta, performed prior to cardiothoracic surgery for the identification of high-risk anatomic features, allows assessment of aortic plaque burden in the entire thoracic aorta in a semiquantitative fashion (7).
We hypothesized that the extent and location of aortic atheroma, determined by pre-operative MDCTA, is associated with long-term mortality following nonaortic cardiothoracic surgery.
We screened the electronic medical record of 10,300 patients who underwent cardiothoracic surgery at our institution between 2002 and 2008. We identified patients who underwent coronary artery bypass graft (CABG) surgery, valvular surgery, or combined bypass and valvular surgery, but excluded patients with concomitant aortic surgery. Out of these, 870 patients had pre-operative contrast-enhanced MDCTA. The studies were performed for evaluation of thoracic and aortic anatomy and/or bypass graft location in the context of surgical planning (7). A total of 8 patients with incomplete clinical data were excluded. Therefore, the final study population included 862 patients. Approval was obtained from the institutional review board at our institution, with waiver of individual consent.
Baseline demographics, past medical history, and medication use were prospectively recorded using the Cardiovascular Information Registry at our institution. Height and weight were measured directly, and body mass index was calculated as weight in kilograms divided by the square of height in meters. Standard demographic data and history of smoking (any prior or current history of smoking), diabetes (patient on insulin or oral hypoglycemic medications, or prescribed a diabetic diet), hypertension (resting blood pressure ≥140/90 mm Hg, or use of antihypertensive medications), coronary artery disease (presence of at least 1 coronary artery stenosis >50% diameter stenosis on coronary angiography), prior myocardial infarction (MI), renal function (glomerular filtration rate [GFR]), peripheral arterial disease (PAD), stroke, and congestive heart failure were recorded. Type of surgery and reoperative status was also recorded.
Primary clinical end points
The primary end point was all-cause mortality during long-term follow-up. Mortality was defined as death of the patient during follow-up and recorded in months after cardiothoracic surgery, based on the Social Security Death Index (8,9).
The pre-operative scans were performed on multiple MDCTA scanner systems (Sensation 16, Sensation 64, and Definition dual-source, Siemens Medical Solutions, Erlangen, Germany; Brilliance 40, Brilliance 64, and iCT 256, Philips Medical, Best, the Netherlands). Spiral imaging with retrospective electrocardiographic gating was performed following the administration of nonionic contrast material via an antecubital vein. The MDCTA protocols for the included indications shared the same craniocaudal coverage (thoracic inlet to diaphragm), but reconstructed slice thickness varied from 1 mm to 3 mm (80% with 3-mm, 15% with 2-mm, and 5% with 1-mm slice thickness). These protocols are associated with estimated effective radiation doses averaging 12 mSv.
Two physicians, blinded to clinical data and surgical history, analyzed the images. Image analysis was performed on a Siemens Leonardo Workstation (Siemens Medical Solutions). Axial images were initially reviewed. Window settings were optimized to best differentiate lumen from vessel wall and minimize calcium blooming artifact. Subsequent measurements were performed in double-oblique cross-sectional images after multiplanar reconstruction along the aortic centerline, using an electronic caliper (Fig. 1,Online Appendix).
The thoracic aorta was divided into 5 segments (Fig. 2): 1) the “Aortic Root,” from the aortic annulus to the sinotubular junction; 2) the “Ascending Aorta,” between the sinotubular junction and the origin of the innominate artery; 3) the “Aortic Arch,” between the innominate artery and including the isthmus; 4) the “Arch Branch Vessel Origins,” defined as the ostia and the proximal 10 mm of the proximal vessel; and 5) the “Descending Aorta,” between the isthmus and the diaphragm.
In each segment, the presence, thickness, circumferential extent, and type of atheroma were evaluated (5,10) (Fig. 3,Table 1). Presence of plaque was defined as visible thickening of the aortic wall. Maximal plaque thickness and the maximal circumferential extent of atheroma were measured in each segment using an electronic caliper. The type of aortic atheromatous plaque was classified as calcified or noncalcified.
Total plaque-burden score
In order to semiquantitatively describe the overall extent and distribution of plaque, we defined a total plaque-burden score (TPBS) by combining plaque thickness and circumferential extent in each of the 5 aortic segments (Table 1).
A score of 1 to 3 was assigned based on plaque thickness (0 = no plaque, 1 = mild [<3 mm], 2 = moderate [3 to 5 mm], and 3 = severe [>5 mm]) (5). Similarly, a score of 1 to 3 was used to describe circumferential plaque extent (0 = no plaque, 1 = mild [<1/3 circumferential diameter], 2 = moderate [1/3 to 2/3 circumferential diameter], and 3 = severe [>2/3 circumferential diameter]) (10). When combined, this resulted in a score of 0 to 6 for each of the 5 segments and, hence, an overall score from 0 to 30.
Interobserver and intraobserver analysis for the TPBS was performed in a subset of 30 randomly chosen cases.
Descriptive statistics for baseline demographics and other clinical variables are recorded as mean ± SD (minimal and maximal value) for continuous variables and percentage for categorical variables. Baseline characteristics were compared among the tertiles of the TPBS using chi-square and Kruskal-Wallis tests (for categorical variables) and 1-way analysis of variance (ANOVA) (for continuous variables).
Thirteen clinical and surgical covariates (including TPBS), considered potential predictors of long-term mortality, were assessed using Cox proportional hazards regression, with Wald chi-square tests assessing the statistical significance of each independent variable (software: PROC PHREG, SAS version 9.2, SAS Institute, Cary, North Carolina). Hazard ratios (HRs) with their associated 95% confidence intervals (CIs) were presented for those predictors found to be significantly associated with long-term mortality. This initial analysis evaluated the association of overall TPBS with mortality, without regard to the distribution of plaque burden among the 5 aortic segments. The relationships between plaque distribution and long-term mortality were specifically analyzed in subsequent, separate Cox models, controlling for the same set of covariates that were included in the initial Cox model that assessed the relationship between overall TPBS and long-term mortality.
Interobserver and intraobserver variability with respect to TPBS were described by the mean difference between the 2 readers and between 2 readings of a single observer. In addition, 95% CIs were reported for the mean difference between readers and between readings within readers. The Bland-Altman limits of agreement, which are generally expected to include 95% of all differences from the same general population, were also calculated (11).
The final statistical analysis included 862 patients. The mean duration of follow-up was 25 ± 16 (range 0 to 80) months.
Baseline demographics, cardiovascular risk factors, and types of surgery of the patient population separated by tertiles of total plaque score are shown in Table 2. There are statistically significant differences of several characteristics among tertiles of plaque score. Specifically, age (p < 0.0001), total cholesterol (p = 0.0003), New York Heart Association functional class (p < 0.0001), frequency of hypertension (p < 0.0001), diabetes (p = 0.0003), smoking (p = 0.0002), PAD (p < 0.0001), history of MI (p < 0.0001), and reoperative surgery (p < 0.0001), increased between tertiles. GFR (p < 0.0001) decreased between tertiles (p values are based on chi-square tests for categorical variables and 1-way ANOVA for continuous variables).
Plaque extent, distribution, and composition
Out of the 862 patients, 127 (15%) had no aortic plaque (TPBS = 0) and 735 (85%) had aortic plaque (TPBS > 0).
Of the 735 patients with aortic plaque (i.e., TPBS > 0), 10 (1.4%) had only noncalcified plaque. The remaining 725 (98.6%) had evidence of calcification somewhere along the thoracic aorta; 647 (88.0%) had only calcified plaque, whereas 78 (10.6%) had both calcified and noncalcified plaque.
Of the 735 patients with plaque, on the basis of maximal plaque thickness, 129 of 735 (17.6%) patients had only mild atheroma (i.e., maximum plaque thickness score of 1 (<3 mm) at 1 or several aortic segments; 306 of 735 (41.6%) patients had moderate atheroma (i.e., maximum plaque thickness score of 2 (3 to 5 mm) at 1 or several aortic segments; and 300 of 735 (40.8%) patients had severe atheroma (i.e., maximum plaque thickness score of 3 (>5 mm) at 1 or several aortic segments).
On the basis of maximal plaque circumferential extent, 334 of 735 (45.4%) patients had only mild atheroma (i.e., maximum circumferential extent score of 1 (<1/3) at 1 or several aortic segments; 271 of 735 (36.9%) patients had moderate atheroma (i.e., maximum circumferential extent score of 2 (1/3 to 2/3) at 1 or several aortic segments; and 130 of 735 (17.7%) patients had severe atheroma (i.e., maximum circumferential extent score of 3 (>2/3) at 1 or several aortic segments).
The mean scores (SD) of plaque thickness in the aortic root, ascending aorta, aortic arch, branch vessels, and descending aorta were 0.46 (±0.69), 0.41 (±0.70), 1.60 (±1.10), 0.87 (±1.00), and 1.45 (±1.05), respectively. The mean scores (SD) of plaque circumferential extent in the aortic root, ascending aorta, aortic arch, branch vessels, and descending aorta were 0.40 (±0.60), 0.35 (±0.56), 1.15 (±0.85), 0.76 (±0.89), and 1.15 (±0.89), respectively (Table 3). The mean TPBS was 8.6 (±6.0) for all patients (Table 3).
Interobserver and intraobserver variability for the TPBS
The interobserver variability for TPBS (n = 29) showed a mean difference between the 2 readers of −1.0 (SD = 2.0), and a median difference of 0 (minimum, maximum differences: −6, +3). Reader 2 tended to assign a higher TPBS than Reader 1 (p = 0.008, signed rank test); however, 70% of all absolute differences were ≤2. The 95% CI for the mean difference between readers in our sample was (−1.76 to −0.24). The Bland-Altman limits of agreement were (−5, +3) (11).
The intraobserver variability for TPBS (n = 30) showed a mean difference between the 2 readings of a single observer to be −0.17 (SD = 1.18), and a median difference of 0 (minimum, maximum differences: −2, +2). The first and second readings were equally likely to yield a higher TPBS (p = 0.52, signed rank test), and 50% of all absolute differences were 1 or less. The 95% CI for the mean difference was (−0.61 to +0.27). The Bland-Altman limits of agreement for intrareader differences were (−2.53, +2.19).
Long-term all-cause mortality
Among all patients, there were a total of 119 confirmed deaths at any time point over the length of the study period. Of these, 29 (24%) were observed in the first 30 days.
Survival rates were examined in tertiles of TPBS (total N = 852, 10 cases excluded due to missing data; first tertile TPBS = 0 to 5 [n = 293], second tertile >5 to 11 [n = 277], and third tertile >11 [n = 282]). Three-year survival estimates based on Product-Limit (Kaplan-Meier) were: Tertile 1 = 94.9% (SE = 0.017), Tertile 2 = 82.8% (SE = 0.028), and Tertile 3 = 75.1% (SE = 0.035). The 3 TPBS tertiles were statistically significantly different with respect to overall survival (p < 0.0001, log-rank test) (Fig. 4).
Relationship of TPBS and long-term mortality
Table 4 shows those predictors significantly associated with long-term mortality in univariate and multivariate analysis.
The TPBS was a statistically significant predictor of mortality (p < 0.0001) while controlling for sex, race, age, BMI, GFR, PAD, hypertension, diabetes mellitus, smoking, history of MI, echocardiographic ejection fraction, reoperative surgery, and surgery type. The estimated HR for TPBS was 1.08 (95% CI: 1.04 to 1.12), indicating that for every 1-unit increase in TPBS, with all other variables fixed, the risk of mortality increased by 8%. Within the 95% CI, the increased risk of mortality corresponding to any 1-U increase in TPBS lies between 4.5% and 12%.
GFR and history of PAD were also statistically significant independent predictors of mortality (p = 0.015 and p = 0.03, respectively), with HRs of 0.989 (95% CI: 0.980 to 0.998) and 1.54 (1.05 to 2.27), respectively. The fourth statistically significant predictor of mortality was type of surgery (p = 0.007). Combined CABG and valvular surgery had the highest risk, followed by valvular surgery alone, and CABG (valvular surgery alone was the reference level in a trichotomous surgery variable).
Influence of plaque location
In order to examine the influence of plaque location, we divided the aorta into an ascending part (root to arch, segments 1 to 4) and a descending part (descending aorta, segment 5). Ninety patients had plaque only in the ascending part, 34 patients had plaque only in the descending part, 611 had plaque in both the ascending and descending part of the aorta, and 127 had no plaque in either portion. We calculated the plaque burden score separately for the ascending part (A-TPBS) and descending part (D-TPBS). The mean (SD) A-TPBS and mean (SD) D-TPBS were 5.97 (±4.62) and 2.60 (±1.88), respectively. There was a strong correlation between the ascending and descending total plaque burden (r = +0.66).
Both A-TPBS (p < 0.0001; HR: 1.10) and D-TPBS (p < 0.002; HR: 1.22) were significant predictors of long-term mortality in the absence of the other (but controlling for the full set of other covariates). When A-TPBS and D-TPBS were both included in a single model (with the other covariates), A-TPBS remained significant (p = 0.0007; HR: 1.09), but D-TPBS lost statistical significance (p = 0.38). In all 3 of these Cox proportional hazards regression models, GFR, PDA, and surgery type were also significant predictors of long-term mortality (p < 0.05), whereas all other covariates were not significant (p > 0.10).
Our data demonstrate an independent relationship between the extent of thoracic aortic atheroma burden and long-term mortality in patients after cardiothoracic surgery. Specifically, a semiquantitative TPBS was a highly significant predictor of long-term mortality, after controlling for baseline demographics, cardiovascular risk factors, and types of surgery (TPBS: p < 0.0001; HR: 1.08). Other independent predictors of mortality were GFR, PAD, and type of surgery.
Previous ultrasound studies in similar patient populations have demonstrated that focal presence and severity of atheroma, of the visualized (ascending) aorta, was a predictor of post-operative adverse events and long-term mortality (3–6). In these studies, the HRs for presence of severe plaque are comparable to our data, previous imaging studies in nonsurgical populations, and clinical data (12,13). Also, similar to previous studies (6), the survival curves in our data (Fig. 4) separate early and increasingly diverge over time, suggesting early and sustained risk with increasing severity of aortic atheroma.
However, previous imaging studies have been limited by the fact that presence and severity of atheroma was based on focal maximal plaque thickness at the most diseased segment of the examined (ascending) aorta. Therefore, the pathophysiologic mechanisms underlying the relationship between atheroma and mortality remains incompletely understood. One explanation, based on the observation of perioperative events, invokes a dominant role of embolic complications, in particular in the context of manipulations of the aorta during surgery. An alternative hypothesis suggests a more systemic effect of the diffuse atherosclerotic disease process, with aortic plaque burden a marker of atherosclerotic disease and complications in other vascular regions.
The semiquantitative description of extent and distribution of atheroma along the entire thoracic aorta in our data provides further insights. Importantly, the relationship of plaque extent to mortality was independent of distribution along the aorta, which is evidence that mortality is related to the systemic atherosclerotic disease process rather than focal (embolic) complications alone. This is further supported by recent studies in the carotid circulation and previous data describing a relationship between coronary and aortic calcification (14,15).
Our data demonstrate that thoracic aortic atherosclerosis is a marker of atherosclerotic events and mortality after cardiothoracic surgery. It is an attractive hypothesis that this relationship is not limited to patients after surgery, but describes general atherosclerotic disease patterns of the aorta, with impact for prevention of cardiovascular events. This requires further evaluation in nonsurgical patient populations.
Recent data from patients evaluated for transcatheter aortic valve implantation, a population with very high mortality, reveal extensive atherosclerotic disease of the aorta (16). On the other extreme of the spectrum, data in population-based asymptomatic individuals, derived from noncontrast-enhanced CT studies, show a relationship between calcified aortic atheroma (aortic calcium scoring) with cardiovascular risk factors (17) and all-cause mortality, independent of conventional cardiovascular disease risk factors (18). However, although in a recent paper both increasing coronary artery calcification (CAC) and thoracic aortic calcification (TAC) were significantly associated with future cardiovascular events, only CAC was an independent predictor, whereas TAC did not add significant predictive power to either the Framingham Risk Score or CAC (15). Comparison of calcified and noncalcified plaque burden is difficult because the relationship between calcified and noncalcified plaque is incompletely understood (19).
Together, these results identify the extent of thoracic atheroma burden as a risk factor for adverse long-term outcome after cardiothoracic surgery. Although our data do not provide definitive evidence, it suggests a relationship to the systemic atherosclerotic disease process and, therefore, has important implications for secondary prevention in post-operative rehabilitation programs. In this context of secondary prevention, it is interesting that previous MDCTA and magnetic resonance imaging studies have shown the potential reversibility of thoracic aortic plaque progression by aggressive risk-factor modification (20,21).
The primary outcome in the current study was all-cause mortality, which was assessed using the Social Security Death Index. Despite limitations, this approach has been shown to be highly specific and unbiased (10,11).
MDCTA has high sensitivity, specificity, and overall accuracy for the identification of severe aortic atheroma (22). However, quantification is limited by calcium blooming artifacts, leading to the overestimation of calcified plaque.
Most patient had calcified plaque, and the number of individuals with only noncalcified plaque of the entire thoracic aorta was low (n = 10), limiting statistical analysis. Within these limitations, our data did not demonstrate a significant interaction between the TPBS and plaque calcification. Due to statistical collinearity between total plaque burdens in the ascending and descending aorta, conclusions about independent role of plaque in the ascending and descending aorta are limited.
There is an established correlation between TAC and CAC. However, coronary calcium scores cannot be reliably determined from contrast-enhanced images, and this information is therefore not available for our patients. Other aortic plaque characteristics and plaque composition may have an impact on outcome (4,14), but assessment is limited with MDCTA.
An inherent selection bias may exist in our data, because performance of a contrast-enhanced MDCTA examination excludes patients with known advanced renal disease. These patients may represent a particularly high-risk group. Also, our patient population included almost exclusively elective surgical indications, because patients that undergo emergent nonaortic surgery typically do not have a prior CT scan.
Our data offer initial “proof of concept” for the use of TPBS as a risk marker for long-term mortality in patients with cardiovascular disease. Further assessment of TPBS should involve several phases, including: prospective validation in independent populations; documentation of incremental information when added to standard risk markers; assessment of impact on patient management and outcomes; and cost effectiveness (23). This is particular true for imaging markers, which are often associated with high cost and, in the case of CT, with radiation exposure (24).
Extent of thoracic aortic atheroma burden is associated with increased long-term mortality in patients following cardiothoracic surgery. These data have implications for pre-operative risk assessment.
For an accompanying slide set, please see the online version of this article.
Dr. Mihaljevic is a consultant for Edwards Lifesciences, St. Jude Medical, and Intuitive Surgical. Dr. Flamm reports indirect departmental research support from Phillips Healthcare and Siemens Medical Solutions. Dr. Halliburton serves on the Medical Advisory Board, Philips Medical Systems. All other authors report they have no relationships to disclose.
- Abbreviations and Acronyms
- total plaque-burden score of the ascending part of the aorta
- coronary artery bypass graft
- coronary artery calcification
- confidence interval
- total plaque-burden score of the descending part of the aorta
- glomerular filtration rate
- hazard ratio
- multidetector-row computed tomographic angiography
- myocardial infarction
- peripheral arterial disease
- thoracic aortic calcification
- total plaque-burden score
- Received July 28, 2010.
- Accepted August 3, 2010.
- American College of Cardiology Foundation
- Eisen A.,
- Tenenbaum A.,
- Koren-Morag N.,
- et al.
- van der Linden J.,
- Hadjinikolaou L.,
- Bergman P.,
- Lindblom D.
- Mitusch R.,
- Doherty C.,
- Wucherpfennig H.,
- et al.
- Davila-Roman V.G.,
- Murphy S.F.,
- Nickerson N.J.,
- Kouchoukos N.T.,
- Schechtman K.B.,
- Barzilai B.
- Thambidorai S.K.,
- Jaffer S.J.,
- Shah T.K.,
- Stewart W.J.,
- Klein A.L.,
- Lauer M.S.
- Boyle C.A.,
- Decoufle P.
- Newman T.B.,
- Brown A.N.
- Sedlis S.P.,
- Morrison D.A.,
- Lorin J.D.,
- et al.
- Hellings W.E.,
- Peeters W.,
- Moll F.L.,
- et al.
- Wong N.D.,
- Gransar H.,
- Shaw L.,
- et al.
- van Werkhoven J.M.,
- Schuijf J.D.,
- Gaemperli O.,
- et al.
- Corti R.,
- Fuster V.,
- Fayad Z.A.,
- et al.
- Hlatky M.A.,
- Greenland P.,
- Arnett D.K.,
- et al.
- Halliburton S.S.,
- Schoenhagen P.