Critical Evaluation of Chest Computed Tomography Scans for Blunt Descending Thoracic Aortic Injury

Transcription

1 Critical Evaluation of Chest Computed Tomography Scans for Blunt Descending Thoracic Aortic Injury Brian A. Bruckner, MD, Daniel J. DiBardino, MD, Todd C. Cumbie, BS, Charles Trinh, MD, Shanda H. Blackmon, MD, Richard G. Fisher, MD, Kenneth L. Mattox, MD, and Mathew J. Wall, MD Michael E. DeBakey Department of Surgery and Department of Radiology, Baylor College of Medicine, Houston, Texas Background. Although aortography has been the longheld gold standard for diagnosis of traumatic blunt aortic injury, advances in imaging technology offer lessinvasive, more-rapid, and potentially more cost-effective evaluation. The purpose of this study was to review this hospital s experience with the screening and diagnosis of blunt aortic injury with emphasis on the critical evaluation of computed tomography (CT) scans for defining descending thoracic aortic injury. Methods. A retrospective single-center analysis of all patients undergoing aortography to evaluate for blunt aortic injury between January 1, 1997, and August 31, 2004, was performed. A policy of relying on CT scans to definitively diagnose blunt aortic injury was not in force, and all patients with positive, equivocal, and negative screening CT scans with significant injury mechanism underwent subsequent aortography; this contributed to an unbiased analysis. A subgroup of patients imaged with the latest generation multislice CT scanners (July 1, 2003, to August 31, 2004) was separately analyzed with rapid three-dimensional reconstruction. Results. Of 856 aortograms, 206 (24.1%) were preceded by chest CT scan. Of 31 patients with confirmed aortic injury, 20 had undergone CT scan with 16 positive for definite injury, 3 positive for possible injury, and 1 false-negative study. Of the 206 patients scanned, 114 (55.3%) showed possible injury, 76 (36.9%) were negative, and 16 (7.8%) were positive. Only 3 of the 114 with possible injury (2.6%) were true positives whereas 1 of the 76 negative scans (1.3%) was a false negative and all 16 positive scans were true positives. These data for CT scan imaging result in a sensitivity of 95%, a specificity of 40%, a positive predictive value of 15%, and a negative predictive value of 99%. Conclusions. Chest CT is an acceptable screening tool based on prerequisite high sensitivity and ease of performance in the trauma patient suspected of having a descending thoracic aortic injury. Although the excellent negative predictive value resulted in an algorithm change at this institution, there were a significant number of equivocal scans that required subsequent aortography. Three-dimensional software reconstruction of the aorta can aid in diagnosing blunt aortic injury when findings are equivocal, but there will continue to be artifacts and limitations that require aortography for clarification. (Ann Thorac Surg 2006;81: ) 2006 by The Society of Thoracic Surgeons Blunt aortic injury (BAI) is second only to head injury as the most common cause of death in blunt trauma [1]. Mortality remains high, with most deaths occurring at the scene of the accident (85%). Those who do survive until presentation at a trauma center often have other associated injuries that can distract from this lifethreatening but treatable entity [2]. With the advent of rapid, high-resolution multislice computed tomographic (CT) scanners, a stable trauma patient can undergo a head, cervical spine, abdomen, and pelvis scan in less than 10 minutes. Because very little time is required to obtain additional slices, patients with significant mechanism of injury or questionable mediastinum by chest radiograph can now undergo screening chest CT scanning in addition to more traditional CT scan imaging. Although several studies have demonstrated that CT scan of the chest is an excellent screening tool for BAI, tremendous controversy surrounds its use as a sole diagnostic modality [3 11]. Computed tomography of the aorta is not without disadvantages and can be affected by technique of study, cardiac and lung movement, volume averaging, inconsistent vascular opacification, and congenital anomalies [7, 12]. Difficulty is encountered when relying on CT scanning for standalone diagnostic imaging, particularly because imaging artifacts or overinterpretation of findings may lead to a high number of false-positive studies and a low positive predictive value. Secondary to this bias, longstanding protocol at the Ben Taub General Hospital calls for aortography to confirm aortic injury after a positive or equivocal screening chest CT. Accepted for publication Nov 3, Address correspondence to Dr Wall, Michael E. DeBakey Department of Surgery, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030; mjwall@houston.rr.com. This article has been selected for the open discussion forum on the CTSNet Web Site: org/sections/newsandviews/discussions/index.html 2006 by The Society of Thoracic Surgeons /06/$32.00 Published by Elsevier Inc doi: /j.athoracsur

2 1340 BRUCKNER ET AL Ann Thorac Surg CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 2006;81: The purpose of this study is to review this institution s experience with the screening and diagnosis of BAI, with emphasis on the critical evaluation of the true predictive value of positive or equivocal chest CT scans. The latest experience with three-dimensional aortic reconstructions is reported, including their limitations. Finally, a practical imaging algorithm is presented that includes CT scanning with three-dimensional reconstruction to aid in the suspected diagnosis of descending aortic injury and relies on aortography for confirmation of injury. Material and Methods All patients who underwent aortogram as a screening modality for BAI from January 1, 1997, to August 31, 2004, were included in the analysis. This study was not submitted for institutional review board approval because individual patient consent was waived and this was a quality review study with no patient identifiers used. During the study period any patient who had a widened mediastinum by chest radiograph or significant mechanism with suspicion of injury was studied either directly by aortogram alone or in combination with screening CT scan. Because of a new technology evaluation and a policy of not relying on CT scans to diagnose BAI, all patients with direct signs of aortic injury by CT (positive study), equivocal CT studies (including mediastinal hematoma), and negative CT studies with significant injury mechanism underwent subsequent aortography. This presented a unique opportunity to evaluate for both false-positive and false-negative scan results and to obtain meaningful data using aortogram as a gold standard. Aortogram was not used selectively, based on the likelihood of injury as read by CT scan. Demographic, imaging, and outcome data such as age, sex, mechanism of injury, chest radiograph findings, CT scan findings, management of aortic injury, operative findings, and outcomes were also collected and stored in database format. Two types of CT scanners were used during the study period; a spiral scanner (January 1, 1997, to September 1, 2003) and a newer generation, multislice helical scanner (July 1, 2003, to August 31, 2004). With aortography being the gold standard for defining injury, the results of the chest radiographs, CT scans, aortograms, and operative findings were compared to determine their relative utility. Technique of Aortography In all cases, the common femoral artery was cannulated using modified Seldinger technique, and a 5F pigtail catheter was advanced to the proximal thoracic ascending aorta. The examination consisted of three digital subtraction angiography runs, in the anteroposterior, left anterior oblique, and steep right anterior oblique right lateral projections. Each run was typically performed with an injection of 20 to 30 ml/s for a total 40 to 50 ml of nonionic contrast, for an average of 120 to 150 ml of contrast used, with images captured at 3 to 6 frames/s. When available, findings of the screening studies (chest radiograph or CT scan) were used by the radiologist in helping to make a final interpretation of the aortogram. Technique of Chest Computed Tomography From January 1, 1997, until June 1, 2003, a General Electric single-slice spiral CT scanner was used in the emergency department at Ben Taub General Hospital. Starting in July of 2003 a high-speed multislice scanner (Somatom Sensation 16 slice, Siemens Medical Solutions, Malvern, PA) was used for all trauma imaging. The protocol for chest CT included a bolus of 100 ml of contrast administered intravenously at a rate of 3 to 4 ml/s. The single slice CT images were acquired with 3.0 to 5.0 mm slice thickness. With the new generation scanners, the images were acquired with 0.75 to 1.5 mm slice collimation, reconstructed as 1.5 to 2.0 mm slice thickness at 1.0 to 1.5 mm intervals for advanced reconstruction and as 3.0 to 5.0 mm slices for axial interpretation. Any abnormalities (ie, mediastinal hematoma, aortic contour irregularities, hemothorax) noted on the axial images prompted immediate two-dimensional and three-dimensional reconstruction using Leonardo workstations running Inspace software and a Syngo environment (Siemens Medical Solutions). Both multiplanar maximal intensity projection and volume rendering technique reconstructed images were reviewed by the most senior radiologist and surgeon. Successive images of a patient with aortic transection are shown in Figure 1 using software reconstruction and eventual subtraction of surrounding anatomic structures to better define the injury. For the purposes of our clinical program and for this study, direct signs of aortic injury include vessel lumen-filling defects, contour abnormality, false aneurysm, or extravasation of contrast. Because the majority of aortic injuries occur in the proximal descending aspect, injuries to the branch vessels, ascending aorta, aortic arch, distal descending aorta, and abdominal aorta were not analyzed and represent a limitation of the current study. Results A total of 856 victims of blunt chest trauma underwent aortogram to evaluate for BAI, with 206 patients (206 of 856, 24%) having undergone a chest CT scan before aortogram and the remaining 650 (650 of 856, 76%) studied directly by means of aortography. All patients underwent a screening chest radiograph as part of the basic trauma workup. Thirty-one patients (31 of 856, 3.6%) were ultimately found to have aortic injury, all located in the proximal descending aorta just distal to the left subclavian artery and all confirmed by aortography. Of the 31 with BAI, 9 were female (9 of 31, 29%) and 22 were male (22 of 31, 71%), with a mean age of 39 years (range, 5 to 70 years). Eighteen of the patients with BAI (18 of 31, 58%) underwent operative repair, and the remaining 13 (13 of 31, 42%) either died before repair (n

3 Ann Thorac Surg BRUCKNER ET AL 2006;81: CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 1341 Fig 1. Successive steps in software three-dimensional reconstruction of an aortic injury. A three-dimensional reconstruction of the patient s thorax is recreated on a computer workstation. Using sophisticated software image subtraction techniques, the surrounding anatomic detail is cleared from the image to reveal the aortic anatomy, and a transection in the descending aorta is seen. 6), were successfully treated medically (n 6), or received a stent graft (n 1). Utility of Chest Radiograph for Blunt Aortic Injury Of the 539 radiographs read as abnormal or indicating a widened mediastinum, 28 had aortic injury (28 of 539, 5%) and 511 did not (511 of 539, 95%). Of the 317 radiographs that were initially read as normal, 3 of these patients (3 of 317, 1%) had BAI and 314 did not (314 of 317, 99%). Of the 31 patients with confirmed BAI, 28 (28 of 31, 90%) had a widened mediastinum by chest radiograph and 3 (3 of 31, 10%) were interpreted as negative for mediastinal hematoma. These data result in a sensitivity of 90%, a specificity of 38% for screening value, a positive predictive value of only 5%, and a negative predictive value of 99% (Table 1). The raw calculations for these values are presented in Appendix 1. Results of Screening Computed Tomographic Scanning for Blunt Aortic Injury A total of 856 aortograms were obtained for blunt chest trauma during the study period, with 206 of these patients (24%) undergoing screening chest CT. At Ben Taub General Hospital the use of chest CT scans for the evaluation of BAI increased dramatically during the study period with a sharp decline in the number of aortograms (Fig 2), such that during the last 2 years of the study nearly all of the patients undergoing aortogram had undergone a prior chest CT scan (Fig 3). Of the 31 patients with BAI, 20 had undergone chest CT, with 16 positive for definite injury (16 of 20, 80%), 3 positive for possible or equivocal injuries (3 of 20, 15%), and 1 demonstrating no injury (1 of 20, 5%). Considering all 206 patients undergoing CT scanning, 114 scans (55%) were read as possible injury, 76 (37%) were read as definitively negative, and 16 (8%) were read as definitively positive. With aortogram as the gold standard, only 3 of 114 with possible injury (2.6%) were true positives, whereas 1 of the 76 negative scans (1.3%) was falsely negative and all 16 positive scans were truly positive. All of the 18 patients who underwent surgery for aortic injury as confirmed by aortogram had intraoperative findings consistent with radiologic diagnosis of injury. These data for CT scan imaging result in a sensitivity of 95%, a specificity of 40%, a positive predictive value of only 15%, and a negative predictive value of 99% (Table 1). Raw calculations for these values are presented in Appendix 2. For the purposes of generating these statistical data, both a definitively positive CT scan reading and a possible injury CT scan reading were considered positive. Using the latest generation CT scanner and threedimensional reconstruction software on the 11 most recent patients with aortic injury, 10 (91%) unequivocally demonstrated descending thoracic aortic injuries, and we believe that these images provide enough anatomic information for operative planning. Additionally, the aver- Table 1. Comparisons Among Imaging Modalities for Patients With Blunt Trauma Undergoing Evaluation for Blunt Aortic Injury Study Number of Studies Sensitivity Specificity PPV NPV Chest radiograph % 38% 5% 99% Chest CT % 40% 15% 99% Aortogram % 100% 100% 100% CT computed tomography; NPV negative predictive value; PPV positive predictive value.

4 1342 BRUCKNER ET AL Ann Thorac Surg CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 2006;81: Fig 2. Total number of aortograms performed for blunt chest trauma per year at the Ben Taub General Hospital from years 1997 through age time from scan to fully reconstructed images took less than 15 minutes when performed by one of two radiologists who had extensive experience with the imaging software. The number of radiologists with expertise in using the workstation software for image reconstruction was a limitation of the study. At the time of this study, more radiologists, including senior residents, were training to use the imaging software so that someone would be available at all times. Figure 4 demonstrates some examples of this technology. Patient A underwent screening CT scan with coronal, sagittal, and threedimensional reconstructions after a high-speed motor vehicle accident. The reconstructed images clearly show a pseudoaneurysm at the aortic isthmus with the aortogram images shown for comparison. Patient B was also involved in high-speed motor vehicle accident, and the reconstructed aortic images demonstrate a long pseudoaneurysm of the proximal descending aorta with the aortogram images shown for comparison. Comment Because of its role in screening for blunt trauma, the chest radiograph remains the first test obtained in evaluation of a patient who may have BAI. It can be performed rapidly in the emergency department, and empiric treatment with -blockade in the stable patient can be initiated. Although there are no prospective data to support its effectiveness, aggressive control of blood pressure and aortic shear forces with -blockade has been shown to be safe. Only 5% of patients with widened mediastinum or abnormal findings on chest radiograph actually had an aortic injury that was read as normal in 10% of patients with BAI. On the basis of these data, the decision to order a screening chest CT should not be made solely on the presence of an abnormal radiograph but also on clinical suspicion, taking into account the injury mechanism. A recent study evaluating the usefulness of chest CT scans by Demetriades and colleagues [4] recommends their routine use irrespective of the roentgenogram findings. However, those patients who have obvious signs of aortic injury by chest radiograph (eg, wide mediastinum greater than 8 cm, significant depression of left mainstem bronchus, nasogastric tube displacement, loss of aortic knob contour) can undergo aortogram after all other indicated CT scans. If the surgeon is comfortable with the radiograph findings, a chest CT scan in these patients may be omitted because it will not change the course of management as an aortogram will be obtained anyway. This study confirms data of numerous other investigators that CT scanning of the chest for BAI is an acceptable screening tool. Only 1 patient with aortic injury had a normal CT scan such that a negative predictive value of 99% and a sensitivity of 95% were obtained, which is ideal for a screening test. The false-negative CT scan was performed in 1997 on the older generation scanner, and the aortogram confirmed a subtle aortic injury that was successfully managed nonoperatively. In the stable patient undergoing transport to the CT scanner for blunt trauma evaluation, it requires very little additional time to obtain a chest scan in addition to the abdominal, head, and other indicated scans with the latest scanner technology, which is also prerequisite for a screening test. Other information, including lung and bony thorax injuries, can be obtained from chest CT. Literature consensus is that if axial images are unremarkable with regard to aortic anatomy and the presence of hematoma, then further workup is not indicated. However, operator experience, institutional experience, and unusual anatomic locations of injury are important limitations of this developing technology. Controversy remains, however, over the role of CT scanning as a definitive study for aortic injury. There are several recent reports that support the role of chest CT scan as a definitive study (thus basing the decision on whether to operating entirely on the scan results) provided that the scan clearly shows injury and is not equivocal, which would probably require an aortogram depending on the practice guidelines at that institution [5, 9, 11]. The specificity of CT scanning for BAI has improved with advances in reconstruction software that can rapidly create coronal, sagittal, oblique planar, and threedimensional images. This was observed when comparing the older and newer generation scanners, and although the sample size is currently small with the new scanners, this trend is expected to continue as the radiologists gain more experience with the imaging software. Fig 3. Total number of aortograms performed for blunt chest trauma per year with percentage undergoing screening computed tomography scan before aortogram (gray bars) from 1997 through 2004.

5 Ann Thorac Surg BRUCKNER ET AL 2006;81: CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 1343 Fig 4. (A) Patient A was involved in highspeed motor vehicle accident. The three-dimensional software reconstructions are compared with the aortograms, demonstrating an identical aortic pseudoaneurysm and anatomy. (B) Patient B was also involved in a high-speed motor vehicle accident. Three-dimensional reconstruction images demonstrate a long pseudoaneurysm involving the proximal descending aorta, identical to that seen on the aortogram (middle image). Patients with positive CT scans can be stratified into two groups: scans that clearly show direct signs of injury (luminal defect, pseudoaneurysm, contrast extravasation) and those that have indirect signs or equivocal findings (mediastinal hematoma, movement artifacts, anatomic anomalies). Stratification into these two groups leads to a significant improvement in the specificity of the positive study regarding scans that show direct signs of injury. Studies that are equivocal are frustrating to the surgeon, and one particularly common finding is mediastinal hematoma without direct signs of aortic injury. Of 74 patients in the current series with mediastinal hematoma as the only significant finding on CT, 3 had a confirmed aortic injury by aortography, a 4% missed injury rate. An earlier study revealed a missed injury rate of 20% in patients with only indirect signs who had aortic injury [3]. Aortography is strongly advocated for patients with any suspicious CT findings, even in the absence of direct signs of injury. Pulsation artifacts from the heart and lungs and suboptimal contrast bolus are two major obstacles for screening chest CT in BAI. Until the image acquiring times become substantially faster or unless the heart is transiently stopped, there will always be some degree

6 1344 BRUCKNER ET AL Ann Thorac Surg CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 2006;81: Fig 5. Problem areas for computed tomographic scanning of blunt aortic injury. (A) Three examples of motion artifact seen in three-dimensional reconstruction of ascending aorta. (B) Two examples of the branch vessels with surrounding structures and artifacts that make identification of injury difficult. of motion artifact (Fig 5A). Because of their smaller size and surrounding structures, branch vessel imaging is also difficult, and clear, unobstructed images are hard to create on the graphics workstation even with extensive software subtraction techniques (Fig 5B). Caution must exist as major limitations of even the current multislice CT scanners are their ability to detect injuries to the ascending aorta, aortic root, or branch vessels. This is a significant problem in evaluating the patient with significant blunt chest trauma, and missed injuries to these structures may occur with overreliance on CT scans. Consider that up to 15% of patients with descending aortic injuries also have injuries to the ascending aorta or branch vessels as determined at autopsy series [13]. Although the majority of imaging artifacts are related to heart and lung motion, other types of artifacts are commonly encountered and commented on in the official scan readings, sometimes resulting in the phrase cannot exclude aortic injury. As in any equivocal study, the surgeon in this situation should obtain an aortogram. There were 40 such patients in this study with unsure reading but without hematoma or direct signs, and one of these patients was confirmed to have aortic injury by aortogram. Another pitfall for the surgeon at hospitals is the initial reading of the study versus the final reading as dictated by the radiologist on call. With all positive CT scans (equivocal and unequivocal studies) undergoing aortography, the aortogram results are available during the final interpretation of the CT scan, which probably contributes a certain amount of bias to these readings. Quality of study issues and adherence to strict protocol are extremely important in generating images for software reconstruction. The site of intravenous access, size of the intravenous catheter, technician experience, and timing of injection all are critical components in CT imaging of the chest. For patients without definite signs of aortic injury on plain radiograph, CT scan of the chest is an acceptable screening tool but cannot be used solely as a definitive study. Although the number of aortograms performed has dramatically decreased across the country and at Ben Taub General Hospital, they are still required for equivocal studies or injuries may be missed. Aortography has several pitfalls including risk of femoral arterial injury and time required to perform study, which can delay treatment. The average wait time for completing an aortogram was 3 hours, which included mobilizing technicians and faculty, and procedure and equipment setup. Therefore, the new imaging technology is welcomed, but one should be cautious about the information obtained and making operative decisions based on these studies. However, using the new imaging technology, direct signs of injury with software reconstruction will probably be the only study needed in the future, with aortograms reserved for equivocal studies.

7 Ann Thorac Surg BRUCKNER ET AL 2006;81: CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 1345 Fig 6. A practical algorithm for suspected blunt aortic injury. *Obvious chest x-ray signs of aortic injury do not need CT chest at this time (if the surgeon is comfortable). Following other CT scans, the stable patient should proceed to aortogram. **At this time, aortogram should be obtained to confirm injury and delineate the operative anatomy. In the future, as technology improves, the surgeon will probably be able to proceed to operation without aortogram. (Abd abdominal; CT computed tomography; 3-D three-dimensional.) A practical algorithm for imaging patients with suspected blunt aortic injury is proposed (Fig 6). After the initial radiograph, stable patients should have screening chest CT if indicated according to mechanism or radiograph findings, and some can proceed directly to aortogram if overt radiographic signs of aortic injury are present. For those patients undergoing screening chest CT scan, if the axial images of the chest are negative, no further workup is necessary for descending thoracic aortic injury, and software image reconstruction is not performed. If the images show any abnormality, software reconstruction should be undertaken by a radiologist experienced with this technology. Clear, unequivocal descending aortic injury, at this time, should still be confirmed by aortogram to precisely define the surrounding anatomy and evaluate for other associated injuries including the ascending aorta and the branch vessels. Computed tomographic scan findings that remain equivocal after reconstruction also require an aortogram to definitively exclude aortic injury. As imaging technology continues to progress in quality and speed, the need for aortogram will continue to decrease but will continue to remain the true, definitive test for years to come. References 1. Fabian TC, Richardson JD, Croce MA, et al. Prospective study of blunt aortic injury: Multicenter Trial of the American Association for the Surgery of Trauma. J Trauma 1997; 42: Wall MJ Jr, Hirshberg A, LeMaire SA, Holcomb J, Mattox K. Thoracic aortic and thoracic vascular injuries. Surg Clin North Am 2001;81:

8 1346 BRUCKNER ET AL Ann Thorac Surg CHEST CT FOR BLUNT DESCENDING THORACIC AORTIC INJURY 2006;81: Chen MY, Miller PR, McLaughlin CA, Kortesis BG, Kavanagh PV, Dyer RB. The trend of using computed tomography in the detection of acute thoracic aortic and branch vessel injury after blunt thoracic trauma: single-center experience over 13 years. J Trauma 2004;56: Demetriades D, Gomez H, Velmahos GC, et al. Routine helical computed tomographic evaluation of the mediastinum in high-risk blunt trauma patients. Arch Surg 1998;133: Downing SW, Sperling JS, Mirvis SE, et al. Experience with spiral computed tomography as the sole diagnostic method for traumatic aortic rupture. Ann Thorac Surg 2001;72: Dyer DS, Moore EE, Ilke DN, et al. Thoracic aortic injury: how predictive is mechanism and is chest computed tomography a reliable screening tool? A prospective study of 1,561 patients. J Trauma 2000;48: Koenig TR, West OC. Diagnosing acute traumatic aortic injury with computed tomography angiography: signs and potential pitfalls. Curr Probl Diagn Radiol 2004;33: Mayberry JC. Imaging in thoracic trauma: the trauma surgeon s perspective. J Thorac Imaging 2000;15: Melton SM, Kerby JD, McGiffin D, et al. The evolution of chest computed tomography for the definitive diagnosis of blunt aortic injury: a single-center experience. J Trauma 2004;56: Mirvis SE, Kostrubiak I, Whitley NO, Goldstein LD, Rodriguez A. Role of CT in excluding major arterial injury after blunt thoracic trauma. AJR Am J Roentgenol 1987; 149: Mirvis SE, Shanmuganathan K, Buell J, Rodriguez A. Use of spiral computed tomography for the assessment of blunt trauma patients with potential aortic injury. J Trauma 1998; 45: Fisher RG, Sanchez-Torres M, Whigham CJ, Thomas JW. Lumps and bumps that mimic acute aortic and brachiocephalic vessel injury. Radiographics 1997;17: Mattox KL. Red River anthology. J Trauma 1997;42: Appendix 1 Statistical Calculations for Chest Radiographic Values for 856 Patients Test Result Positive Negative totals Positive a 28 b 3 31 Disease Presence Negative c 511 d Totals Sensitivity a/(a b) 28/31 90% Specificity d/(c d) 314/ % Positive predictive value a/(a c) 28/539 5% Negative predictive value d/(d b) 314/317 99% Appendix 2 Statistical Calculations for Chest Computed Tomographic Values in 206 Patients Test Result Positive Negative totals Positive a 19 b 1 20 Disease Presence Negative c 111 d Totals Sensitivity a/(a b) 19/20 95% Specificity d/(c d) 75/186 40% Positive predictive value a/(a c) 19/130 15% Negative predictive value d/(d b) 75/76 99% INVITED COMMENTARY The authors [1] report a single center experience of 856 aortograms performed to evaluate suspected descending thoracic aortic injury from 1997 to In 206 patients, a computed tomographic (CT) examination preceded the aortogram, using a single-slice CT scanner until 2003 and a modern multidetector (16-slice) CT scanner thereafter. Any such study faces challenges (ie, maintaining a unified protocol despite rapid technical advances in imaging, encompassing a variety of clinical styles and levels of experience, accommodating the evolving paradigm and diagnostic demands of endovascular treatment of traumatic rupture, and accumulating enough patients in a reasonably discrete period when the prevalence of the index injury is less than 4%). The authors cannot hope to make everyone happy. In the end, the authors limit their conclusions to two: (1) that chest computed tomography is an acceptable screening tool with high sensitivity, and (2) that 3-dimensional reconstruction tools are helpful but not conclusive for detecting injury. The authors acknowledge issues that qualify even the following modest claims: clearing the descending thoracic aorta in isolation has limited utility as injuries may be multiple, whereas artifacts and reader inexperience may limit 3-dimensional reformatting and other advanced CT techniques. Regarding the latter, our department and others have reported a less than 1% incidence of significant deviation between resident and staff readings of emergency CT scans. Furthermore, the importance of immediate definitive CT interpretation may be overemphasized, as many aortic ruptures are now treated in delayed fashion. Does this study apply a fixed diagnostic algorithm to answer a specific clinical question in a well-defined population? Figure 3 documents a decline in aortography from a high of 180 in 1997 to a low of 40 in Computed tomographic scans have been steady at 30 to 40 per year. Thirty-one aortic injuries were found, 11 of them in the current era of the multidetector CT and 3-dimensional reconstruction. These 11 (30%) injuries were found in the latest 14 months (July 1, 2003 to August 31, 2004) using approximately 50 aortograms in contrast to 20 injuries (70%) in the preceding 78 months using 800 aortograms. Using multidetector computed tomography, we see a monthly incidence of aortic rupture three times 2006 by The Society of Thoracic Surgeons /06/$32.00 Published by Elsevier Inc doi: /j.athoracsur