Summary receiver operating characteristic (ROC) curve analysis of transesophageal echocardiography (TEE), computed tomography (CT), and magnetic resonance imaging (MRI). The weighted summary ROC curve is expressed by a solid line. Individual study estimates of sensitivity and (1 – specificity) are shown by open circles. Solid squares indicate pooled point estimates of sensitivity and specificity.
Posttest probability according to pretest probability. Posttest probability was calculated as follows: Posttest probability = [pretest odds × likelihood ratio]/[(1+pretest odds) × likelihood ratio]; where pretest odds = pretest probability/(1 – pretest probability). CT indicates computed tomography; MRI, magnetic resonance imaging; and TEE, transesophageal echocardiography.
Shiga T, Wajima Z, Apfel CC, Inoue T, Ohe Y. Diagnostic Accuracy of Transesophageal Echocardiography, Helical Computed Tomography, and Magnetic Resonance Imaging for Suspected Thoracic Aortic DissectionSystematic Review and Meta-analysis. Arch Intern Med. 2006;166(13):1350-1356. doi:10.1001/archinte.166.13.1350
Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
Patients with suspected thoracic aortic dissection require early and accurate diagnosis. Aortography has been replaced by less invasive imaging techniques including transesophageal echocardiography (TEE), helical computed tomography (CT), and magnetic resonance imaging (MRI); however, accuracies have varied from trial to trial, and which imaging technique should be applied to which risk population remains unclear. We systematically reviewed the diagnostic accuracy of these imaging techniques in patients with suspected thoracic aortic dissection.
Published English-language reports on the diagnosis of thoracic aortic dissection by TEE, helical CT, or MRI were identified from electronic databases. Sensitivity, specificity, and positive and negative likelihood ratios were pooled in a random-effects model.
Sixteen studies involving a total of 1139 patients were selected. Pooled sensitivity (98%-100%) and specificity (95%-98%) were comparable between imaging techniques. The pooled positive likelihood ratio appeared to be higher for MRI (positive likelihood ratio, 25.3; 95% confidence interval, 11.1-57.1) than for TEE (14.1; 6.0-33.2) or helical CT (13.9; 4.2-46.0). If a patient had shown a 50% pretest probability of thoracic aortic dissection (high risk), he or she had a 93% to 96% posttest probability of thoracic aortic dissection following a positive result of each imaging test. If a patient had a 5% pretest probability of thoracic aortic dissection (low risk), he or she had a 0.1% to 0.3% posttest probability of thoracic aortic dissection following a negative result of each imaging test.
All 3 imaging techniques, ie, TEE, helical CT, and MRI, yield clinically equally reliable diagnostic values for confirming or ruling out thoracic aortic dissection.
The mortality rate associated with thoracic aortic dissection is high1 and has recently been reported to rise by 1% to 1.4% per hour when a patient remains untreated, leading to a 68% mortality rate within 48 hours.2,3 The mortality rate is highest in patients with type A dissection (involving the ascending aorta) and has been reported to be 58% without but still 26% with surgical treatment.1 Conversely, the mortality rate for patients with type B dissection is lower, but more importantly, patients treated medically have an 11% lower mortality rate than those treated surgically, for which it is reported to be 31%.1 Therefore, early and accurate diagnosis and decision making regarding surgical or conservative intervention are essential to reduce morbidity and mortality among patients with clinically suspected thoracic aortic dissection.
Beginning in the 1960s, aortography was used as a standard tool for assessing patients with clinically suspected thoracic aortic dissection. However, the technique is invasive, costly, potentially nephrotoxic owing to contrast materials and ionized radiation, and time consuming, sometimes causing diagnostic delays.2 More importantly, the diagnostic accuracy of aortography is not as high as originally thought. According to a European cooperative study, sensitivity and specificity for the diagnosis of aortic dissection are 88% and 94%, respectively.4
Over the last 2 decades, aortography has been used less frequently, as noninvasive imaging techniques including transesophageal echocardiography (TEE), helical computed tomography (CT), or magnetic resonance imaging (MRI) have emerged. Use of these imaging techniques tends to be based on availability at local hospitals, the degree of emergency, or whether they can be used in combination with another technique,3 but cumulative data on the diagnostic accuracy of each of these techniques remain limited and have varied from trial to trial. In addition, which imaging technique should be applied to which risk population remains unclear. We, therefore, have systematically reviewed the diagnostic accuracy of each of these imaging techniques in patients with suspected thoracic aortic dissection.
We searched MEDLINE (January 1980 through August 2005) and the Cochrane Library (2005, issue 3) for reports of studies and trials related to the method used to diagnose thoracic aortic dissection. Only English-language articles were included. The initial search terms were “thoracic aortic dissection,” “transesophageal echocardiography (TEE),” “helical CT” (or “spiral CT”), and “magnetic resonance imaging (MRI).” A manual search of REFERENCES cited in published reports and reviews was also performed.
Reports were independently selected and reviewed by 2 investigators (T.S. and Z.W.). The systematic review process for selection of eligible studies is shown in Figure 1. Reported studies were selected if they met the following criteria: (1) the study was prospective; (2) at least 1 imaging technique was used; (3) absolute numbers of true-positive, false-negative, true-negative, and false-positive results were available or could be derived from the published data; and (4) the reference standard for diagnosing thoracic aortic dissection was clearly indicated. We excluded retrospective studies, studies with insufficient data, and studies that did not focus on thoracic aortic dissection but rather focused on thoracic aortic disease in general.
Studies were graded for quality according to the modification of a priori criteria, as described by Romagnuolo et al5: (1) blinding; (2) consecutive recruitment of patients; and (3) single (vs composite) reference standards.
We defined the presence of aortic dissection as the presence of 2 vascular lumens separated by an intimal flap within the aorta, as in most studies included in the review. We defined surgical, autopsy, or angiographic findings as the reference standard for thoracic aortic dissection, as in most studies included in the review. Extracted from the reports were the number of patients, mean age, general patient characteristics, diagnostic criteria for aortic dissection, the reference standard, onset and types of dissection, system-specific settings, and absolute numbers of true-positive, false-positive, true-negative, and false-negative results.
We calculated pooled estimates of sensitivity, specificity, positive and negative likelihood ratios, and the natural logarithm of the diagnostic odds ratio by the DerSimonian-Laird random-effects model.6 Rates were pooled after logit transformation, weighting study rates by the inverse ratio of their variance plus the between-study variance for that measure, and then retransformed back into standard proportions with 95% confidence intervals (CIs). Homogeneity of the effect size across trials was tested by χ2 statistics. Heterogeneity was defined as P<.10.
The diagnostic performance of each test was also assessed by means of summary receiver operating characteristic (ROC) curves according to the method described by Moses et al.7 In construction of summary ROC curves, the true-positive rate was plotted against the false-positive rate for each study. To avoid calculation problems by having 0 values, 0.5 was added to each cell of the respective contingency table. The summary ROC model is described by the following equation: D = a + bS. The summary ROC curve analysis is based on regression analysis of logistic regression–transformed data, which plots the difference between the logistic regression of the true-positive rate (TPR) and that of the false-positive rate (FPR) (D = logit TPR – logit FPR) on the y-axis and their sum (S = logit TPR + logit FPR) on the x-axis. The y axis (D) is equivalent to the log diagnostic odds ratio, and the x-axis (S) is a measure of how the test characteristics vary with the test threshold. The regression coefficient b examines the extent to which the log odds ratio is dependent on the threshold values chosen. The linear regression analysis was weighted by the inverse of the variance of D. The regression line was back-transformed to the ROC space.
To assess the potential for publication bias, a funnel plot was constructed in which the log of relative risks was plotted against the associated number of patients. In addition, correlation between the standardized log of relative risks and the associated number of patients was determined by the Kendall rank correlation coefficient.8 The correlation between sample size and relative risk would be strong if not many small studies with null results were published. A significant correlation between sample size and relative risk would not exist in the absence of publication bias. Statistical significance was defined for treatment effects as P<.05, and heterogeneity and publication bias were assumed when P<.10. Analyses were performed with Microsoft Excel (Microsoft Corporation, Redmond, Wash), Meta-DiSc Version 1.2 for Windows (Hospital Ramón y Cajal, Madrid, Spain), and Number Cruncher Statistical System 2004 (NCSS Statistical Systems, Kaysville, Utah).
The electronic search resulted in 90 hits. Sixteen studies9- 24 representing a total of 1139 patients met the inclusion criteria (Table 1 and Table 2). Pooled estimates of sensitivity, specificity, positive and negative likelihood ratios, and the natural logarithm of the diagnostic odds ratio as well as the regression model equation for each test are listed in Table 3. The positive likelihood ratio was highest for MRI, suggesting a possibly superior discriminative power for confirming thoracic aortic dissection. Helical CT had the lowest negative likelihood ratio, suggesting that it might be best for ruling out thoracic aortic dissection. The highest diagnostic odds ratio suggests that of the 3 imaging techniques, MRI might be superior in overall diagnostic performance. The summary ROC curve for each test is shown in Figure 2. Diagnostic accuracy did not vary with the test threshold for TEE (P = .45), helical CT (P = .53), or MRI (P = .53). This suggests that the odds ratio is independent of the threshold values chosen.
The posttest probabilities following positive and negative test results are depicted with a possible range of pretest probabilities in Figure 3. According to Sarasin and colleagues,3 in terms of diagnostic probability, patients with a pretest probability of 50% for aortic dissection should be considered high-risk patients, and patients with a 5% pretest probability should be considered low-risk patients. High-risk patients were defined as patients with typical clinical findings of aortic dissection (eg, tearing chest pain in the back, hypotension, and disparity in pulses). Low-risk patients were defined as patients with severe and acute chest pain and no other suggestive vascular or neurologic findings. In our results, if a patient had a 50% pretest probability of thoracic aortic dissection, he or she had a 96%, 93%, or 93% posttest probability of thoracic aortic dissection following a positive result for MRI, TEE, or helical CT, respectively. In contrast, if a patient had a 5% pretest probability of thoracic aortic dissection (low-risk population), he or she had a 0.1%, 0.2%, or 0.3% posttest probability of thoracic aortic dissection following a negative result for helical CT, TEE, or MRI, respectively.
Symmetry in the funnel plot was confirmed by a significant Kendall correlation coefficient of 0.11 (P = .65) for TEE, 1.0 (P = .11) for helical CT, and 0.14 (P = .65) for MRI, suggesting absence of publication bias.
In the present meta-analysis, we found that sensitivity and specificity were comparable between imaging techniques, and the diagnostic value of each imaging technique was acceptable for confirming or ruling out thoracic aortic dissection. In patients at high risk for thoracic aortic dissection (pretest probability of 50%), MRI yielded the highest values for confirming thoracic aortic dissection. In contrast, helical CT yielded the best values for ruling out thoracic aortic dissection in patients at low risk for thoracic aortic dissection (pretest probability of 5%). However, capabilities of the 3 imaging techniques were found to be equivalent for confirming or excluding thoracic aortic dissection. It is therefore difficult to determine from our results which imaging technique should be recommended after stratification of patients according to pretest probabilities.
Transesophageal echocardiography is often used in the primary care setting. Compared with MRI and helical CT, TEE is advantageous in emergency situations with time constraints and in patients with hemodynamic compromise. The amount of time needed for diagnosis may be shortest with TEE; this seems very important because the progress of complications is time dependent. Color-flow Doppler TEE is advantageous in confirming the presence of aortic valve insufficiency associated with aortic dissection. However, the distal part of the ascending aorta and the branches of the aortic arch cannot be adequately evaluated by TEE.25 Interference by the trachea and the left main stem bronchus produces a “blind zone” or “blind spot.”2 Generally, TEE examination tends to be somewhat observer and experience dependent, and false-positive or false-negative findings can arise easily when data are interpreted by unskilled or single observers.26 In patients with esophageal varices, TEE is contraindicated.
Currently, conventional CT is probably the most widely used imaging technique in the diagnosis of thoracic aortic dissection,1 but it is reported to be associated with an insufficient sensitivity of 83% to 94% and a specificity of 87% to 100%.2,12 The next generation of imaging technology, helical CT, offers many advantages over conventional CT: temporal resolution is improved; motion artifacts are minimal; scanning is more rapid; examination time is substantially shorter; and 3-dimensional renderings are possible. Our analysis showed that helical CT yielded a sensitivity of 100% and a specificity of 98% and, of the 3 techniques, the best value for ruling out thoracic aortic dissection when a patient is at low risk for thoracic aortic dissection (pretest probability of 5%). Of the 3 techniques, CT may be the least operator dependent. The disadvantages of helical CT include the need for contrast material or ionizing radiation. This may be a problem especially when the patient is in a state of acute or chronic renal compromise. Aortic valve insufficiency is difficult to evaluate with helical CT.
Magnetic resonance imaging has been considered the most accurate technique for diagnosing thoracic aortic dissection.2 Our meta-analysis confirmed that the overall diagnostic accuracy for detecting thoracic aortic dissection is excellent. Magnetic resonance imaging also yielded the best value for confirming thoracic aortic dissection when a patient is at high risk for this disorder (pretest probability of 50%). The results were homogeneous, irrespective of the type of MRI study, such as cine magnetic resonance angiography.10 Despite the advantages, MRI is rarely used as the initial imaging technique1 because of lack of availability, time delay, incompatibility with implanted metal devices, or monitoring difficulties during examination. Magnetic resonance imaging is not applicable for hemodynamically unstable patients.
Our analysis showed that the diagnostic odds ratio is independent of the test threshold chosen. This is likely because most studies defined aortic dissection as the presence of 2 vascular lumens separated by an intimal flap. Some studies included an additional criterion for aortic dissection: central displacement of intimal calcification was considered aortic dissection if thrombosis of the false lumen was present; however, this additional criterion did not affect the overall diagnostic performance of the 3 imaging techniques. With the exception of a few parameters in TEE and helical CT results, heterogeneity did not exist between studies.
Despite the strengths of our meta-analysis, our findings are limited by the following. First, our systematic review did not permit head-to-head comparisons of the 3 imaging techniques because only 3 studies12,18,20 directly compared 2 or more imaging techniques in the same cohort. Second, there were several criteria included in the reference standard (eg, surgical findings and autopsy findings). This may be inevitable because a single reference standard would not be applicable in the same cohort; surgery is performed in some cases and not in others, and the same is true for autopsy, angiography, and other procedures. Third, some of our findings are limited by the low number of studies meeting our inclusion criteria, especially those of helical CT. This might explain why the summary ROC curve for spiral CT looks irregular. Finally, in addition to confirming or ruling out thoracic aortic dissection, there are other important factors that must be evaluated: the presence of branch vessel or coronary artery involvement, and the presence of aortic valve insufficiency. These factors strongly influence the type of surgical intervention. However, our meta-analysis could not determine the role of any of the 3 imaging techniques in this respect because few studies evaluated this.
Given the equally high performance of each test, what other information should a clinician use to choose the most appropriate test for a given patient? A clinician must take into account the availability of each imaging test because time delay increases the mortality rate in untreated patients.3 A clinician also needs to recognize that there is wide disparity in levels of expertise and resources at local hospitals. A clinician should not hesitate to order another test when thoracic aortic dissection is still clinically strongly suspected but the initial diagnostic test is negative. This was clearly recommended in the report based on the International Registry of Acute Aortic Dissection27; however, delay must be avoided.
In conclusion, our systematic review suggests that all 3 imaging techniques, ie, TEE, helical CT, and MRI, yield equally reliable diagnostic values for confirming or ruling out thoracic aortic dissection.
Correspondence: Toshiya Shiga, MD, PhD, Department of Anesthesiology, Toho University Ohashi Medical Center, Ohashi 2-17-6, Meguro-ku, Tokyo 153-8515, Japan (QZX02115@nifty.com).
Accepted for Publication: April 10, 2006.
Financial Disclosure: None reported.