Association of Internal Carotid Artery Injury with Carotid Canal Fractures in Patients with Head

Neuroradiology York et al. CT of Internal Carotid Artery Injury Gerald York 1,2 Daniel Barboriak 1 Jeffrey Petrella 1 David DeLong 1 James M. Provenzale 1 York G, Barboriak D, Petrella J, Delong D, Provenzale J Received April 30, 2004; accepted after revision September 9, 2004. 1 Department of Radiology, Box 3808, Duke University Medical Center, Durham, NC 27710. Address correspondence to J. M. Provenzale. 2 Present address: Serving with the United States Army in Iraq. AJR 2005;184:1672 1678 0361 803X/05/1845 1672 American Roentgen Ray Society Association of Internal Carotid Artery Injury with Carotid Canal Fractures in Patients with Head Trauma OBJECTIVE. The purpose of our study was to determine the degree to which carotid canal fracture and other CT findings are associated with internal carotid artery (ICA) injury in patients with head trauma. MATERIALS AND METHODS. Three neuroradiologists retrospectively evaluated CT scans and cerebral angiograms of 43 patients who underwent cerebral angiography within 7 days after blunt cranial trauma over a 5-year period. Seventeen patients underwent unilateral and 26 had bilateral carotid angiography. Angiograms were evaluated for ICA injury and CT scans were evaluated for carotid canal fracture, brain contusion, subarachnoid hemorrhage, basilar skull fracture, subdural hematoma, soft-tissue swelling, sphenoid sinus air fluid level, and other skull fracture. We recorded the number of true-positive (+CT, +angiogram), truenegative ( CT, angiogram), false-positive (+CT, angiogram), and false-negative ( CT, +angiogram) studies. We determined the sensitivity, specificity, positive predictive value, and negative predictive value for each CT finding. RESULTS. We identified 21 carotid canal fractures in 17 patients. Eleven ICA injuries were seen in 10 patients. Six patients with ICA injury had a carotid canal fracture. The presence of a carotid canal fracture had a sensitivity of 60% and specificity of 67% for detection of injury to the ICA passing through that canal. These values were similar to those for other CT findings. CONCLUSION. Sensitivity, specificity, positive predictive value, and negative predictive value of carotid canal fracture were only moderately good for determining the presence of ICA injury and were similar to other CT findings not typically associated with ICA injury. arotid artery injury is reported to C occur in approximately 1% of individuals who experience severe blunt head trauma [1, 2]. Complications of carotid injury can include stroke, pseudoaneurysm formation, and death. In one study, permanent neurologic damage was seen in nearly 40% of patients and death in approximately 30% [3]. Given the potentially serious consequences of an undiagnosed vascular injury, screening criteria that are primarily based on clinical features have been proposed and implemented by several trauma centers to increase the sensitivity of detecting craniocervical vascular injury [1, 2]. Because many patients with blunt cranial trauma undergo CT, determination of CT findings that might serve as predictors of carotid artery injury would be valuable. One recent study found that the risk of carotid artery injury was substantially greater in patients with carotid canal fracture than in patients with other base-of-skull fractures who did not have carotid canal fracture [4]. However, that study did not address the central issue of the likelihood of carotid injury in patients with carotid canal fracture. To our knowledge, no articles in the medical literature address this important issue or provide details of the frequency with which carotid canal fracture and other CT findings are associated with internal carotid artery (ICA) injury. In this study, we attempted to determine the degree to which carotid canal fracture is associated with ICA injury. Our hypotheses were that the sensitivity, specificity, positive predictive value, and negative predictive value of carotid canal fracture for detecting ICA injury would be substantially better than for other CT findings [1] and that the frequency of ICA injury would be substantially greater in patients with carotid canal fracture than in other patients evaluated for traumatic ICA injury [2]. Materials and Methods We reviewed a handwritten angiography log book at our university-based level one trauma cen- 1672 AJR:184, May 2005

CT of Internal Carotid Artery Injury A B Fig. 1. True-positive CT finding for internal carotid artery (ICA) injury in 51-year-old man who was ejected from automobile during motor vehicle accident. A, Unenhanced axial CT scan using soft-tissue reconstruction with wide window setting shows bilateral carotid canal fractures (arrows). B, Catheter angiogram, left common carotid artery injection, lateral view, shows small pseudoaneurysm (arrow) consistent with arterial dissection. Right common carotid artery injection (not shown) also showed ICA dissection, producing two true-positive findings. ter for the names of patients listed as having undergone carotid angiography for head trauma during the period 1992 1996. The initial review showed the names of 61 patients. Twelve patients with penetrating cervical or cranial injury were excluded. The names of the remaining 49 patients were then cross-referenced with the electronic medical records at our institution to determine which patients had undergone CT and cerebral angiography within 1 week after head trauma. The 43 patients who met this criterion formed our study population. The CT and angiography reports of all patients were reviewed in the electronic medical record to confirm that the studies were performed for assessment of traumatic injury. This review provided the mechanism of injury in 35 patients, which included motor vehicle accident in 24 patients, a fall in seven patients, and assault in four patients. The mechanism of trauma could not be determined in eight patients. The final interpretations of CT reports and angiography reports were recorded. Three CAQ-certified neuroradiologists performed a review of CT scans and angiograms. At the time of image review, reviewers were not blinded to patient name or to the fact that patients had experienced trauma. However, they were blinded to the CT scan and angiogram interpretations. One of the reviewers was a member of the seven-person neuroradiology team at the time of clinical interpretation of angiograms and CT scans and could potentially have provided the official interpretation for four of the CT scans and two of the angiograms. However, because of the long interval (varying between 3 and 7 years) between the time of official CT and angiogram interpretations and the time of blinded interpretations, reviewer bias due to knowledge of clinical interpretations was deemed to be low. Two of the reviewers joined the neuroradiology team after the angiograms and CT scans were obtained and therefore did not provide any official interpretations or studies. All carotid angiograms were performed using digital subtraction angiography and the Seldinger technique; the angiograms always provided anteroposterior and lateral views using common carotid artery injections of contrast material centered at the level of the skull base. Six angiograms had an additional oblique view of the carotid artery. In all cases, the carotid artery was visible from a few centimeters below the carotid bifurcation to the supraclinoid segment of the ICA. Sixteen patients underwent unilateral carotid angiography and 27 patients, bilateral angiography, resulting in 70 carotid arteries that could be assessed. Reviewers independently reviewed filmed (i.e., hard-copy) images from the angiograms. Reviewers were aware that patients had sustained cranial trauma but were blinded to more specific clinical history and the previous interpretation of angiography reports. These reviewers recorded whether ICA injury was present or absent. The diagnosis of ICA injury was made when one or more of the following criteria were met: pseudoaneurysm formation, focal region of luminal narrowing, or presence of an intimal flap. Thereafter, the reviewers met within 2 weeks of completion of the independent review and decided by consensus whether ICA injury was present or absent. Although it was not the goal of this study to assess the rate of correct interpretation of angiograms, consensus interpretations of angiograms were compared with official reports to explain why in some cases angiography might have been performed when an ipsilateral carotid canal fracture was not identified on the consensus interpretation (see following text). A total of 58 CT scans were obtained in the 43 patients within 1 week after head trauma. Fifty of these CT scans were available for this study. Fractures in a given patient were counted only once. Unenhanced CT was performed using either 5-mm (39 CT scans in 32 patients) or 3-mm (11 CT scans in 11 patients) contiguous axial images from the level of the foramen magnum to the vertex. One of four third-generation CT scanners (9800 series, GE Healthcare) was used. Therefore, a total of 11 patients underwent CT using 3-mm collimation and 32 patients underwent CT with solely 5-mm collimation. The same three observers who reviewed angiograms also independently reviewed all 50 CT scans. As with angiograms, consensus decisions on AJR:184, May 2005 1673

York et al. A Fig. 2. False-negative finding for internal carotid artery (ICA) injury in 32-year-old woman involved in motor vehicle accident. A, Unenhanced axial CT scan using soft-tissue reconstruction with wide window setting shows no evidence of carotid canal fracture. Adjacent images (not shown) also showed no evidence of fracture. B, Catheter angiogram, right common carotid artery injection, lateral view, shows abrupt tapering of contrast column in upper cervical ICA consistent with dissection, indicating false-negative finding. CT scan findings were compared with official reports to understand why in some cases angiography might have been performed when an ipsilateral carotid canal fracture was not identified on the consensus interpretation. The most common explanation was that the official report stated a carotid canal fracture was present on the side on which angiography was performed. To decrease the likelihood that reviewers would recall angiography findings for a specific patient when evaluating CT scans, the CT scan review was performed between 6 and 10 weeks (average, 8 weeks) after the initial review of angiograms. Reviewers assessed both carotid canals in all 43 patients, for a total of 86 carotid canals. Because a total of 70 carotid arteries were studied in these patients, we could correlate the presence or absence of carotid canal fracture with the presence or absence of carotid artery injury in 70 instances. All CT scans were reviewed using soft-tissue window settings and softtissue reconstruction with a wide window setting because filmed bone algorithm reconstruction images were not routinely available. All three observers independently reviewed the filmed images of the bone window and soft-tissue window settings from the CT scans and decided whether one or more of the following features were present or absent: fracture of one or both carotid canals, brain contusion, subarachnoid hemorrhage, basilar skull fracture, subdural hematoma, soft-tissue swelling, air fluid level in the sphenoid sinus, and a nonbasilar skull fracture. Because nasotracheal intubation can produce an air fluid level as a result of retained secretions rather than traumatic hemorrhage, this feature was evaluated only on CT scans in which the patient was not receiving intubation. The determinations were then compared and a decision was made by consensus. We recorded the number of true-positive (+CT, +angiogram), true-negative ( CT, angiogram), false-positive (+CT, angiogram), and false-negative ( CT, +angiogram) studies using each CT finding. We then determined the sensitivity, specificity, positive predictive value, and negative predictive value for each CT finding. Also, a Pearson chi-square test of association of ICA injury with each CT finding was performed, with statistical significance set at a p value of less than 0.05, and confidence intervals for the proportions were computed from the binomial distribution (rather than from the asymptotic normal approximation). These confidence intervals are based on the methods of Clopper and Pearson [5] and are sometimes called exact confidence intervals. Logistic regression for pairs of variables was also performed to assess improvement in prediction that might occur by the use of more than a single variable. We used pairs of CT findings, rather than all CT findings in a single logistic regression, to jointly predict ICA injury because the data set was too small and the data were too sparse to allow the reasonable use of the logistic model with all predictors simultaneously. The fact that both a right carotid canal fracture and a left carotid canal fracture were examined for association with ICA injury was taken into account by using a Bonferroni adjustment for this pair of findings. In addition, because seven CT findings were simultaneously considered for association with ICA injury, we also used a Bonferroni adjustment for this set of findings. B 1674 AJR:184, May 2005

CT of Internal Carotid Artery Injury TABLE 1 CT Findings in Patients with Carotid Canal Fractures, Including Presence or Absence of Fracture with Internal Carotid Artery Injury Results Basilar Skull Fracture Results ICA Injuries On the consensus review of angiograms, reviewers found seven instances of right ICA injury and four instances of left ICA injury in 10 patients (10 dissections, one carotid cavernous fistula) or 23% of our study population (Fig. 1). All ICA injuries except one (Fig. 2) were within 2 cm of the carotid canal fracture. Six patients with carotid canal fracture (35% of patients with carotid canal fracture) had an ICA injury (including one who had bilateral ICA dissections and a unilateral fracture), and four patients with no carotid canal fracture had an ICA injury. Four patients with an ICA injury did not have a carotid canal fracture, representing 15% of patients without a carotid canal fracture (Fig. 2). Therefore, six patients with ICA injury (or 60% of all patients with ICA injury) had a carotid canal fracture. One patient had bilateral ICA injuries with a unilateral carotid canal fracture. One patient with right ICA injury underwent only right (not bilateral) common carotid angiography. All remaining patients with ICA injury underwent bilateral angiography. Eleven patients who had a unilateral carotid canal facture underwent bilateral angiography, and one of these patients had an ICA injury. Carotid Canal Fractures On the consensus review of CT scans, reviewers identified 21 carotid canal fractures in 17 patients (bilateral fractures in four patients, unilateral fractures in 13 patients). Nine patients had solely right carotid canal Contusion Subdural Hematoma CT Finding Subarachnoid Hemorrhage fracture, and four patients had solely left carotid canal fracture. A fracture was not seen in 65 carotid canals. Four (24%) of the 17 patients with carotid canal fracture had bilateral fractures. Distribution of unilateral or bilateral angiography among patients according to the presence of carotid canal fracture was as follows: Bilateral ICA angiography was performed in 14 of 26 patients with no carotid canal fracture, nine of 13 patients with a single carotid canal fracture, and two of four patients with bilateral carotid canal fractures (Fig. 1). The remaining patients underwent unilateral angiography. In the three of the four patients with a single carotid canal fracture who underwent unilateral angiography, the angiography was performed ipsilateral to the side of the fracture. In the single patient with a consensus interpretation of bilateral carotid canal fractures who underwent unilateral angiography, the official interpretation was that of a unilateral fracture on the side on which angiography was performed, which likely accounts for the reason only unilateral angiography was performed. In all, 17 disagreements occurred between official interpretations and consensus interpretations of CT scans. These included 14 carotid canals in which the original interpretation was that of fracture but which, on consensus review, were determined to not have a fracture. All 14 patients had an angiogram ipsilateral to the carotid canal that did not have a fracture on the consensus interpretation, and in all cases the angiogram was normal. Three carotid canal fractures (in three patients) were Soft-Tissue Swelling Air Fluid Level Skull Fracture a No. of patients having finding 17 (40%) 11 (26%) 12 (28%) 15 (35%) 30 (70%) 24 (56%) 17 (40%) (% of all patients) + ICA injury 4 4 3 3 5 6 4 + CC fracture + ICA injury 3 1 1 3 3 4 1 CC fracture ICA injury 7 4 4 7 9 9 7 + CC fracture ICA injury 15 10 14 12 30 18 16 CC fracture No angiography corresponding to one CC 5 3 2 5 13 11 6 Note. Numbers in top row are numbers of patients with specific CT finding. Numbers in remaining rows are numbers of carotid canals. Because each patient has two carotid canals, numbers in top row equal half the sum of numbers in column below them. ICA = internal carotid artery, CC = carotid canal, + denotes present, denotes absent. a All skull fractures other than basilar skull fracture and carotid canal fractures. interpreted as normal on the official interpretation but as having a fracture on the consensus interpretation. In one of these patients, a negative angiogram was obtained on the side of the carotid canal fracture identified on the consensus interpretation. In two other patients, angiography was performed solely on the side contralateral to the fracture. This fact explains why, although a total of 19 fractures were identified on the consensus interpretations, two fractures did not have a corresponding angiogram. Therefore, although 19 carotid canal fractures were detected on the consensus interpretation, our number of truepositive cases (carotid canal angiogram pairs that showed both a fracture and an ICA injury) and false-positive cases (carotid canal angiogram pairs that showed a fracture but a normal angiogram) is 17 rather than 19. The rates of various combinations of positive and negative findings for carotid canal fracture and ICA injury are listed in Table 1. Basilar skull fracture was present in 17 patients (40% of all patients), of whom six (35% of these patients) had a total of seven ICA injuries. Four (15%) of the 26 patients without basilar skull fractures were found to have ICA injuries. Cerebral contusion was seen in 11 patients (26% of all patients), of whom five (45%) had ICA injuries. Five (16%) of the 32 patients who did not have contusions were found to have ICA injuries. Reviewers found subdural hematomas in 12 patients (28% of all patients), of whom four (33%) had ICA injuries. Six (19%) of the 31 patients who did not have subdural hematomas had ICA injuries. Reviewers found AJR:184, May 2005 1675

York et al. TABLE 2 Performance Measure Relationship of Various CT Findings to Presence of Internal Carotid Artery Injury Carotid Canal Fracture a Sensitivity 60% (0.262 0.878%) Specificity 67% (0.482 0.820%) PPV 35% (0.142 0.617%) NPV 85% (0.651 0.956%) Basilar Skull Fracture 60% (0.262 0.878%) 67% (0.482 0.820%) 35% (0.142 0.617%) 85% (0.651 0.956%) Note. 95% confidence intervals are indicated in parentheses. PPV = positive predictive value, NPV = negative predictive value. a Presence of a carotid canal fracture with either an ipsilateral or contralateral internal carotid artery injury. b All skull fractures other than basilar skull fracture and carotid canal fractures. subarachnoid hemorrhage in 15 patients (35% of all patients), of whom six (40%) had ICA injuries. Four (14%) of the 28 patients without subarachnoid hemorrhage had ICA injuries. Scalp soft-tissue swelling was seen in 30 patients (70% of all patients), of whom eight (27%) had ICA injuries. Two (15%) of the 13 patients who did not have scalp soft-tissue swelling were found to have ICA injuries. Reviewers found an air fluid level in the sphenoid sinus in 24 patients (56% of all patients), nine (38%) of whom had a total of 10 ICA injuries. One (5%) of the 19 patients who did not have an air fluid level in the sphenoid sinus was found to have an ICA injury. Reviewers found a nonbasilar skull fracture in 17 patients (40% of all patients), of whom five (29%) had ICA injuries. Five (19%) of the 26 patients who did not have a nonbasilar skull fracture were found to have a total of six ICA injuries. Contusion 50% (0.187 0.813%) 82% (0.645 0.930%) 46% (0.167 0.766%) 84% (0.672 0.947%) Subdural Hematoma 40% (0.122 0.738%) 76% (0.577 0.889%) 33% (0.099 0.651%) 81% (0.625 0.925%) Subarachnoid Hemorrhage 60% (0.262 0.878%) 73% (0.545 0.867%) 40% (0.163 0.677%) 86% (0.673 0.960%) Association of CT Findings with Presence of ICA Injury The frequency with which CT findings were associated with ICA injury and confidence intervals are listed in Table 2. The presence of a carotid canal fracture had a sensitivity of 60% and a specificity of 67%, positive predictive value of 35%, and negative predictive value of 85% for detecting an injury to the ICA that coursed through that carotid canal. The sensitivity and specificity of carotid canal fracture were similar to those for many other CT findings that are not typically associated with traumatic ICA injury. Highest association of other CT findings with ICA injury is described in the following text. A sphenoid sinus air fluid level had the highest sensitivity (90%), followed by soft-tissue swelling (80%), compared with 60% for carotid canal fracture. Contusion (82%) and subdural hematoma (76%) were the two features with the highest specificity, compared with 67% for carotid canal fracture. Contusion and subarachnoid hemorrhage had the highest positive predictive values (46% and 40%, respectively). A sphenoid sinus air fluid level had the highest negative predictive value (95%). The positive and negative predictive values for carotid canal fracture were similar to those for other CT findings. Chi-square values for correlation of ICA injury (on any side) with CT findings were as follows: left carotid canal fracture, 0.02 (p > 0.0.2); right carotid canal fracture, 5.47 (p = 0.04); basilar skull fracture, 2.28 (p > 0.0.2); subdural hematoma, 0.95 (p > 0.0.2); subarachnoid hemorrhage, 3.62 (p > 0.0.2); soft-tissue swelling, 0.65 (p > 0.0.2); air fluid level in the sphenoid sinus, 6.17 (p = 0.1); and other skull fracture, 0.60 (p > 0.0.2). The analysis of pairs of variables did not show a statistically significant relationship between any pair of CT findings and ICA injury. None of the joint logistic models indicated any statistically significant or useful additional predictive value obtained by using pairs of findings. Soft Tissue Swelling 80% (0.444 0.975%) 33% (0.180 0.518%) 27% (0.123 0.459%) 85% (0.546 0.981%) Air Fluid Level 90% (0.555 0.997%) 55% (0.364 0.719%) 38% (0.188 0.594%) 95% (0.740 0.999%) Skull Fracture b 50% (0.187 0.813%) 64% (0.451 0.796%) 29% (0.103 0.560%) 81% (0.606 0.934%) Discussion Previous studies have noted the significance of carotid canal fracture in relationship to ICA injury [2, 6]. However, little information regarding the frequency of ICA injury after carotid canal fracture is available in the medical literature [4]. In this study, we set out to test two hypotheses: First, that the sensitivity, specificity, positive predictive value, and negative predictive value of carotid canal fracture for detecting ICA injury would be substantially better than for other CT findings. Second, that the frequency of ICA injury would be substantially higher in patients with carotid canal fracture than in other patients evaluated for traumatic ICA injury. The frequency of ICA injury among patients with carotid canal fractures (35%) in our study was more than twice as high as that of patients who did not have a carotid canal fracture (15%). Although this difference did not meet statistical significance in our small study, such an association is plausible on the basis of our data. These findings suggest that carotid canal fracture is an indication to proceed to further evaluation of the carotid artery using catheter angiography or a noninvasive technique such as CT angiography, Doppler sonography, or MR angiography. On the basis of our data, individual physicians must decide for themselves the tolerable thresholds for acceptance of false-negative studies in this population of patients. A future study evaluating the risk benefit ratio of imaging trauma patients would be worthwhile, especially in an era when noninvasive tests for evaluating arterial injury have gained increased acceptance. Overall, the association of carotid canal fracture and ICA injury was relatively equivalent, but not superior, to that seen for other CT findings. The presence of a sphenoid sinus air fluid level had a much greater sensitivity (90%) than carotid canal fracture (60%). The specificity of 67% for carotid canal fracture was less than that for cerebral contusion (82%) and subdural hematoma (76%) and about the same as that for basilar skull fracture (67%), other skull fracture (64%), and subarachnoid hemorrhage (72%). No single CT finding clearly had superiority with regard to a combination of sensitivity, specificity, positive predictive value, or negative predictive value relative to other CT findings. However, positive and negative predictive values are highly prevalence-dependent and therefore only generalizeable to patient populations similar to the one studied here. 1676 AJR:184, May 2005

CT of Internal Carotid Artery Injury Statistical analysis of association of any single CT finding with ICA injury failed to show any statistically meaningful association. Although the association of right carotid canal fracture and ICA injury exceeded the formal significance level, the same level of significance was not seen on the left side, and no anatomic reason exists for a difference between sides. The highest associations were with an air fluid level in the sphenoid sinus (p = 0.1) and subarachnoid hemorrhage (p > 0.0.2). In addition, no pair of CT findings showed any statistically significant improvement of association with ICA injury over single CT findings. ICA injury was found in two of six patients with basilar skull fracture who did not have carotid canal fracture, which was approximately the same rate of ICA injury as in patients with carotid canal fractures (35%). One previous study reported finding 11% of patients with carotid canal fracture related carotid injury compared with 2% of patients with basilar skull fracture but no carotid canal fracture [4]. Compared with that study, we found a higher rate of ICA injury both in patients with carotid canal fractures and in those without fractures. We also found no substantial difference in rates of ICA injury between those two patient groups (rather than a fivefold increase in ICA injury rate among patients with carotid canal fractures). In this study, a determination of the presence of carotid canal fracture was made by a retrospective review of CT scans. In a substantial number of cases, the consensus determination made by the reviewers differed from that rendered at the time of clinical interpretation, which explains why some patients who did not have a carotid canal fracture underwent ICA angiography. However, the frequency of ICA injury in those patients was not trivial (15%). Future prospective studies that take into account factors other than CT findings, such as mechanism of injury, neurologic status, and presence of other, non-cns injuries, are needed to provide information that will guide physicians as to which patients who do not have carotid canal fracture should undergo vascular assessment. The decision to perform bilateral ICA angiography when a single (rather than bilateral) carotid canal fracture is seen is one that angiographers who deal with trauma patients must consider. Our study provides some information to help with this decision. We found 11 patients who had a unilateral carotid canal facture in whom bilateral angiography was performed. One of these patients had an ICA injury contralateral to the side of a unilateral fracture. Individual angiographers may consider that the 9% frequency of a contralateral ICA injury is worth the risks of bilateral angiography in a patient who is already undergoing angiography for one carotid canal fracture. The development of noninvasive techniques for detecting ICA injury may markedly influence the decision as to whether catheter angiography should be considered at all in patients with carotid canal fracture. The 24% frequency of bilateral carotid canal fractures seen in our study among patients with a carotid canal fracture is similar to that of 27% reported in a previous study of carotid canal fractures [4]. In individual patients in our series, the likelihood of ICA injury increased with the number of carotid canal fractures. Half of the four patients with bilateral carotid canal fractures had an ICA injury compared with the frequency of 23% in the overall study population, 31% in patients with single carotid canal fracture, and 15% in patients with no carotid canal fracture. However, the small number of patients in these subgroups prevents any meaningful comparison. Possibly the increased likelihood of ICA injury was only indirectly related to the number of fractures and was instead related to other features, such as mechanism of injury or increased degree of trauma, which independently increase the likelihood of ICA injury. Because we had only limited information about the mechanisms of injury in our patients, we were not able to answer such questions. Given the opinion of many physicians that presence of a carotid canal fracture is an indication for cerebral angiography or other assessment of the ICA, one might expect that rate of carotid canal fracture would be much greater in patients with ICA injury than in patients who do not have ICA injury. In fact, the rate of at least one carotid canal fracture in patients with ICA injury (60%) was indeed higher than in patients who did not have ICA injury (33%). The difference was not statistically significant between the two groups, which is likely a function of the small sample. Among the 10 patients who had ICA injuries, the frequencies of various carotid canal fracture combinations were similar to one another: two patients had bilateral carotid canal fractures, four patients had one carotid canal fracture, and four patients had no carotid canal fracture. A number of limitations are evident in our study. First, CT was performed at two levels of collimation, 3 mm and 5 mm. The use of a relatively thick collimation of 5 mm in approximately two thirds of patients decreased sensitivity and specificity for detection of carotid canal fractures. Specifically, the number of carotid canal fractures may be underestimated; perhaps that is one reason for the number of carotid artery injuries without carotid canal fractures. Future studies using a uniform technique with thin collimation are indicated to verify our findings. Second, our study may have underestimated the number of ICA injuries because approximately one third of patients did not undergo bilateral carotid angiography. A third limitation is that we had limited access to information about mechanism of trauma and to information governing the decision process for choice of angiography in particular, choice of unilateral as opposed to bilateral angiography. Nonetheless, detailed information about specific mechanisms of injury in many common types of trauma (e.g., motor vehicle accidents) is often unavailable to treating physicians even at the time patients are being evaluated, or multiple mechanisms are operative. Therefore, the amount of detailed information about mechanism of trauma in many of the patients in our study may be close to that which could be obtained in a prospective study. Next, as is expected in any retrospective study, information as to the preliminary interpretation of a CT study (which may have guided the decision to perform angiography and the side on which angiography was to be performed), as opposed to the official interpretation of a CT study, is lacking. In addition, the decision processes underlying the performance of angiography are lacking. Unilateral angiography may have been performed in some cases because the angiographer did not consider it necessary to assess both carotid arteries in the setting of a solitary carotid canal fracture. In other cases, angiography may have been terminated in some cases after solely unilateral angiography because the presence of a positive finding was considered sufficient for clinical purposes. This lack of detailed information about the reasons for angiography limits the extent to which our data can be interpreted. Another limitation is inherent in the fact that we performed a consensus review of CT scans. Although this procedure has the advantage of a systematic review by well-qualified neuroradiologists, it understandably resulted in discrepancies between official interpretations at the time of scanning and the consensus interpreta- AJR:184, May 2005 1677

York et al. tion, which produces limitations on the validity of interpretation of our results. An additional limitation is that some element of selection bias exists on a number of levels because a small number of carotid canal or ICA injuries may have not been detected. For instance, some patients who had carotid canal fracture may not have undergone CT or angiography. Because patients who did not undergo both CT and angiography at our institution were excluded, a patient who underwent one or both of those studies at another institution would have escaped our detection. However, we expect that the number of such patients was low. Furthermore, reviewers were not blinded to the clinical indication for CT or angiography, which may have influenced image interpretation. Another limitation of our study is the relatively small number of patients. However, this series represents almost all patients who underwent catheter angiography in the posttrauma setting over a 6-year period at a busy level 1 trauma center. Although a larger number of patients might have provided more definitive data, a study period substantially longer than our 6-year period of observation could prove difficult. Finally, we did not use a group of patients without trauma and with normal angiography and CT findings to control for observer bias. Ultimately, the decision whether to perform a test on a patient depends on a number of factors, including the pretest probability of a finding, the degree of invasiveness of a test, and the relative benefit risk ratio for the patient. In this study, we show that the frequency of ICA injury associated with carotid canal fracture in our small population was high, but that the sensitivity and specificity of carotid canal fracture for predicting ICA injury were not better than those of many other CT findings that are not typically thought to be predictive. During the period in which our patients were evaluated, confidence in MR angiography among many referring clinicians was likely lower than at present. CT angiography was not realistically an option during the time period in which our patients were assessed. As confidence in noninvasive techniques such as MR angiography and CT angiography increases, the cost risk ratio for assessing ICA injury also increases. Our study provides valuable data for beginning an understanding of the frequency of ICA injury in patients with carotid canal fracture; future studies that are prospective are needed. As noninvasive tests for ICA injury become more accepted and the risk to patients becomes even lower, the threshold for evaluating arterial injury in these patients will also be lowered. The importance of the information gained from a noninvasive examination will likely outweigh the small likelihood of traumatic ICA injury in certain subsets of this population. References 1. Biffl WL, Moore EE, Ryu RK, et al. The unrecognized epidemic of blunt carotid arterial injuries: early diagnosis improves neurologic outcome. Ann Surg 1998;228:462 470 2. Kerwin AJ, Bynoe RP, Murray J, et al. Liberalized screening for blunt carotid and vertebral artery injuries is justified. J Trauma 2001;51: 308 314 3. Fabian TC, Patton JH, Croce MA, et al. Blunt carotid injury: importance of early diagnosis and anticoagulant therapy. Ann Surg 1996;223:513 525 4. Resnick DK, Subach BR, Marion DW. The significance of carotid canal involvement in basilar cranial fracture. Neurosurgery 1997;40:1177 1181 5. Clopper CJ, Pearson ES. The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika 1934;26:404 413 6. Carter DA, Mehelas TJ, Savolaine ER, Dougherty LS. Basal skull fracture with traumatic polycranial neuropathy and occluded left carotid artery: significance of fractures along the course of the carotid artery. J Trauma 1998;44:230 235 1678 AJR:184, May 2005