cute oral or neck bleeding in patients

Transcription

1 Neuroradiology/Head and Neck Imaging Original Research Ku et al. Antecedent CT or MRI for Prediction of Hemorrhage Sites Neuroradiology/Head and Neck Imaging Original Research Yi-Kang Ku 1 Yon-Cheong Wong 1 Chen-Ju Fu 1 Hsiao-Jung Tseng 2 Li-Jen Wang 1 Chao-Jan Wang 1 Shy-Chyi Chin 1 Ku YK, Wong WC, Fu CJ, et al. Keywords: carotid blowout syndrome, common carotid artery, downstream external carotid artery, internal carotid artery, proximal external carotid artery, transarterial embolization DOI: /AJR Received July 28, 2015; accepted after revision November 12, Supported by grant CMRPG from Chang Gung Memorial Hospital at Linkou, School of Medicine, Chang Gung University. 1 Department of Medical Imaging and Intervention, Chang Gung Memorial Hospital at Linkou, School of Medicine, Chang Gung University, No. 5 Fu Sing Rd, Kweishan, Taoyuan 333, Taiwan, ROC. Address correspondence to S. C. Chin (b25chin@gmail.com). 2 Biostatistical Center for Clinical Research, Chang Gung Memorial Hospital at Linkou, School of Medicine, Chang Gung University, Taoyuan, Taiwan. Supplemental Data Available online at AJR 2016; 206: X/16/ American Roentgen Ray Society Timely Antecedent CT or MRI Can Help Predict Hemorrhage Site of Posttreatment Head and Neck Cancer, With Digital Subtraction Angiography Used as the Reference Standard OBJECTIVE. We investigated the timing of CT and MRI performed before digital subtraction angiography (DSA) in the prediction of hemorrhage sites in patients with head and neck cancers who present with acute oral or neck bleeding after receiving treatment. MATERIALS AND METHODS. A total of 123 DSA examinations that evaluated 123 oral or neck bleeding events in 85 patients were analyzed. The last CT or MRI examinations performed within a time frame of days before transarterial embolization were reviewed retrospectively, with three findings (pseudoaneurysm, air-containing necrotic tissue, and residual tumor) used to predict hemorrhage sites. DSA findings of pseudoaneurysm or active contrast extravasation were used as a reference standard. The sensitivity of CT and MRI for correctly predicting hemorrhage sites was used to determine the optimal timing of CT or MRI examinations performed before DSA. RESULTS. A total of 8.9% of the DSA examinations (11/123) had equivocal findings but were followed by another bleeding event for which DSA findings were positive. CT or MRI was statistically significantly better at predicting hemorrhage sites in patients with bleeding events associated with nonhypopharyngeal cancers (p = 0.019) than in those with bleeding events associated with hypopharyngeal cancers. The sensitivity of CT or MRI in the prediction of hemorrhage sites was statistically significantly higher for the common carotid artery and the internal carotid artery when CT or MRI was performed less than 30 days before bleeding events occurred. Prediction of hemorrhagic sites was better with the use of CT angiography than with the use of enhanced CT or MRI, although it was not statistically significant. CONCLUSION. DSA findings can temporarily be equivocal. CT or MRI examinations performed within 30 days of bleeding events can predict the site of hemorrhage. If no CT or MRI findings from the past 30 days are available, we suggest performing emergent CT angiography for the sake of obtaining better arterial detail. A cute oral or neck bleeding in patients with head and neck cancers is often referred to as carotid blowout syndrome and can be life threatening. This syndrome is frequently discussed in the field of head and neck oncology, but previous studies have primarily focused on associated digital subtraction angiography (DSA) findings and endovascular management [1, 2]. When urgent endovascular intervention is requested, knowing whether the hemorrhage site is from the external carotid artery (ECA), the internal carotid artery (ICA), or the common carotid artery (CCA) has profound implications for the type of treatment selected (i.e., sacrifice vs vessel preservation). In this setting, results of an antecedent CT or MRI examination can offer valuable information for identifying hemor- rhage sites and facilitating treatment, especially for the occasional case for which DSA results are equivocal. A similar observation has been noted by other researchers [3, 4]. In the present study, we sought to further assess how the optimal timing of an antecedent CT or MRI examination can influence prediction of hemorrhage sites. Materials and Methods Patients and the Timing of Imaging Studies This retrospective study included patients with posttreatment head and neck cancers complicated by acute oral or neck bleeding during the 10 years between January 2004 and December All patients underwent emergent DSA because they had significant bleeding accompanied by unstable blood pressure. A total of 93 patients were initially enrolled in the study, but eight patients were ex- AJR:206, April

2 Ku et al. cluded because results of their CT or MRI examinations were unavailable or were suboptimal because of previously implanted coils or stents. All patients had a history of head and neck cancer in the oral cavity, nasopharynx, oropharynx, or hypopharynx, and they all underwent surgery and chemoradiation, usually within 2 years of diagnosis. All patients who were enrolled in the study underwent posttreatment surveillance CT or MRI, followed by DSA. The interval between the time that CT or MRI was last performed and the time that DSA was performed was expressed as the number of days. The demographic and clinical characteristics of the patients are shown in Table 1. Definition of Cancer Sites and the Anatomy of Cervical Arteries and Head and Neck Structures The primary sites of and treatment strategy for head and neck cancers were determined on the basis of pathologic findings and the consensus diagnosis determined by a multidisciplinary head and neck cancer board before treatment. Given that the treatment strategy for bleeding in the proximal ECA is similar to that for bleeding in the CCA and the ICA, we used the beginning point of the faciolingual branch as a landmark for dividing the ECA into the proximal ECA and the downstream ECA. We also defined other arteries as bleeding arteries from the carotid system contralateral to the primary cancer or bleeding arteries not from the carotid system, such as vertebral thyrocervical arteries. Our definitions and descriptions of the rest of the head and neck structures and the vascular anatomy, such as the CCA and the ICA, are the same as commonly used terms. CT and MRI Protocols Because of the retrospective design of this study, a uniform imaging protocol could not be achieved. Most contrast-enhanced CT and CT angiography (CTA) studies were performed using a 16-MDCT scanner (Somatom Sensation 16, Siemens Healthcare). In general, CT images were obtained using an FOV of 280 mm and a matrix size of Most MRI studies were performed using a 1.5-T system (Gyroscan Intera, Philips Healthcare) with the use of a head and neck coil and the following imaging parameters: an FOV of mm, a matrix size of , and T1-weighted spin-echo with a TR/TE of 400/15. T2-weighted turbo spin-echo imaging was performed using a TR/TE of 4000/120. The longitudinal scanning field for both CT and MRI studies was from the upper chest to the skull base that included the aortic arch, the CCA, the ICA, the ECA, and the vertebral arteries. A slice thickness of 4 mm with a 1-mm interval was used for the MRI study, and a reconstructed slice TABLE 1: Demographic and Clinical Characteristics of 85 Patients With Carotid Blowout Syndrome Who Underwent Transarterial Embolization (TAE) Demographic or Clinical Characteristic Value Age (y) Mean ± SD 52 ± 8.7 Range Sex Male 79 (93) Female 6 (7) Location of diagnosed cancer Nasopharynx 21 (24.7) Oropharynx 11 (12.9) Hypopharynx and larynx 14 (16.5) Oral cavity 39 (45.9) Digital subtraction angiography No. of examinations performed 123 Examinations with positive findings 112 (91.1) Examinations with negative findings 11 (8.9) No. of single examinations performed 64 No. of repeat examinations performed 48 Interval between last CT or MRI study performed before TAE (d) Mean ± SD 44 ± 57 Range Last CT or MRI studies performed before TAE (no.) CT 87 MRI 25 All 112 Finding from last CT/MRI performed before TAE Residual tumor 69 (61.6) Air-containing necrotic tissue 90 (80.4) Pseudoaneurysm 21 (18.8) Location of hemorrhage sites, as determined by DSA Common carotid artery 23 (20.5) Internal carotid artery 26 (23.2) External carotid artery Proximal ECA (near bifurcation) 6 (5.4) Downstream ECA 53 (47.3) Locations of hemorrhage sites predicted on CT or MRI Common carotid artery 22 (19.6) Internal carotid artery 24 (21.4) External carotid artery Proximal ECA (near bifurcation) 3 (2.7) Downstream ECA 57 (50.9) Uncertain 6 (5.4) Note Except where otherwise indicated, data are no. (%) of patients. DSA = digital subtraction angiography, ECA = external carotid artery. 830 AJR:206, April 2016

3 Antecedent CT or MRI for Prediction of Hemorrhage Sites thickness of 3 mm was used for the enhanced CT study in the axial, coronal, and sagittal planes. The amount of IV contrast agent administered was ml of iohexol (350 mg I/mL; Omnipaque 350, GE Healthcare), for CT or CTA studies, or gadopentetate dimeglumine (0.2 ml/kg of body weight; Magnevist, Bayer), for MRI studies. For CTA studies, the rate of contrast medium injection was ml/s, and the 1- to 2-mm reformatted axial images of the early arterial phase were used for detection of hemorrhagic sites. Digital Subtraction Angiography Protocol The DSA protocol used for emergency cases of carotid blowout syndrome was usually as simple as possible. Via a femoral approach, angiograms of the bilateral CCA, ECA, and ICA were acquired. Angiograms of the vertebral artery and the subclavian artery were acquired if carotid angiograms revealed negative findings. If a hemorrhage site was identified by the diagnostic DSA study, an endovascular procedure was performed to stop the bleeding. On the basis of the bleeding site(s), the vital signs of the patient, and the preference of the treatment interventionist, different treatments could be chosen, including embolization of the bleeding site, sacrifice of the parent artery, or preservation of the parent artery with the use of a cover stent. Interpretation of Images Of importance, before we reviewed the DSA findings, we determined the most likely hemorrhage sites on the basis of CT or MRI findings. Two radiologists independently reviewed each CT or MRI study and each DSA study, while blinded to the other results, and a third radiologist, who was also blinded to the results, served as referee if the interpretations were discordant. Interpretation of CT and MR Images for Prediction of the Hemorrhage Site Many findings have been suggested as signs predictive of the hemorrhage site, including arterial wall enhancement, luminal compression and irregularity, pseudoaneurysm formation, air-containing necrotic tissue, and residual tumor [4 6]. To determine which CT or MRI findings are present when episodes of bleeding occur, we examined the imaging findings for 18 patients who underwent contrast-enhanced CT or MRI within 1 day of DSA, which was performed at the time of hemorrhage. We found that pseudoaneurysm, air-containing necrotic tissue, and residual tumor were present in 14 (77.8%), 15 (83.3%), and 8 (44.4%) of the 18 patients, respectively. These three findings were used for prediction of the hemorrhage site in the present study because the two radiologists had greater consensus for these findings compared Fig. 1 Comparison of sensitivity of CT and MRI examinations for prediction of hemorrhage site on basis of interval of 30 days or less versus more than 30 days between time that CT or MRI was performed and time of bleeding event. Bar graph shows statistically significant differences (asterisks) in sensitivity for prediction of hemorrhage sites in internal carotid artery (ICA) (p = ) and common carotid artery (CCA) (p = 0.007). Note that, for prediction of proximal external carotid artery (ECA) as hemorrhage site, no CT or MRI study performed more than 30 days from time of bleeding event was available. Sensitivity (%) 100 with other imaging findings. In general, MRI is better than CT for identifying tissue necrosis and residual tumor, whereas contrast-enhanced CT is better than MRI for the detection of air bubbles, pseudoaneurysm, and active extravasation. Pseudoaneurysm A pseudoaneurysm is defined by the presence of a saccular or fusiform lesion that is commonly accompanied by extravasation of contrast agent when active bleeding occurs. Air-containing necrotic tissue After chemoradiation therapy, tissue necrosis is an advanced result of inflammation and a sign of ongoing neck infection. Imaging findings commonly consist of mucosal ulceration and air-containing necrotic tissue, with or without involvement of the facial or * * ICA Proximal ECA 30 Days Downstream ECA > 30 Days CCA skull bones. Our definition of ongoing infection is similar to the term air-containing pharyngocutaneous fistula, as defined in a previously published article [3]; however, it more broadly includes infection in the skull base, oral cavity, and lateral neck. Tissue necrosis in the lateral neck may injure the proximal ECA, the CCA, or the ICA, whereas tissue necrosis in the masticator space and floor of the mouth most likely causes hemorrhage from branches of the downstream ECA. Tissue necrosis in the skull base can affect the bilateral ICA and the downstream ECA branches. An area with no enhancement can indicate fibrosis. If not mixed with air bubbles, an unenhanced area cannot be considered to be tissue necrosis [6]. A B Fig year-old man with hypopharyngeal cancer (posttreatment) with three bleeding events in left carotid system. A and B, CT scans acquired 1 day before first bleeding event showed extensive air-containing necrotic tissue in left hypopharynx (arrow, A) and carotid space (arrow, B). Image from first digital subtraction angiography (DSA) (not shown) revealed pseudoaneurysm from left common carotid artery (CCA) that was treated with covered stent. (Fig. 2 continues on next page) AJR:206, April

4 Ku et al. C D Fig. 2 (continued) 59-year-old man with hypopharyngeal cancer (posttreatment) with three bleeding events in left carotid system. C and D, DSA image of second bleeding event, obtained 31 days after CT was performed, shows small pseudoaneurysm from left superior thyroid artery (arrow, C). Fifty days later, DSA image of third bleeding event shows another pseudoaneurysm near proximal end of stented CCA lumen (bracket, D). When the downstream ECA, proximal ECA, ICA, or CCA was in similar proximity to a region of air-containing necrotic tissue, the ICA, CCA, or proximal ECA was preferentially chosen for the ensuing endovascular treatment after consideration of the risk of brain stroke associated with the hemorrhage sites. When air-containing necrotic tissue was noted in the masticator space or the floor of the mouth, far from the carotid space, the downstream ECA was considered to be the hemorrhage site, despite the fact that small branches of the ECA sometimes are not seen on CT or MR images. Residual tumor Posttreatment CT or MR images identified varying tumor statuses, such as complete or partial tumor regression. The tumor sometimes was not easily outlined in the presence of accompanying inflammation and osteonecrosis, so we generally considered any masslike enhancing lesion to be a residual tumor. The presence of an air-containing masslike lesion is considered to be suggestive of the coexistence of residual tumor and tissue necrosis. Interpretation of Digital Subtraction Angiography Images There is a well-known classification system for DSA findings for use in the evaluation of the severity of blunt carotid arterial injury [7]. However, we consider mild intimal injury, luminal narrowing, and even increased vascular blush as equivocal DSA findings for which endovascular treatment is not justified. The bleeding site is definitive only when there is clear visualization of active contrast extravasation or a pseudoaneurysm. In the patient population with treated head and neck cancer, increased vascular blush is commonly seen on DSA images without hemorrhage. Statistical Analysis The kappa value was used to evaluate agreement between the two radiologists with regard to detection of the same bleeding events. The chisquare test or Fisher exact test was applied to measure three comparisons: the relationship between the primary cancer and prediction of the hemorrhage site by CT or MRI, the consensus between respective carotid arteries and prediction of the hemorrhage site by CT or MRI, and differences among contrast-enhanced CT, CTA, and MRI in the prediction of the hemorrhage site. The Mantel-Haenszel linear association test was used to detect a trend for the correct prediction by three CT or MRI signs. An ROC curve was drawn on the basis of the interval (expressed in days) between the time the last CT or MRI examination was performed and the time DSA was performed, for predicting the consensus on hemorrhage sites. The optimal timing of CT or MRI for reliable prediction of the hemorrhage site before DSA was performed was determined using the Youden index. The sensitivities and specificities of CT or MRI prediction of the hemorrhage sites for the respective arteries were then calculated on the basis of the optimal cutoff point of the time interval. All statistical analyses were performed using statistical software (SPSS software, version 17.0, SPSS-IBM). Statistical significance was denoted by p < Results Characteristics at Baseline A total of 123 DSA examinations that evaluated 123 bleeding events in 85 patients were analyzed. Eleven DSA examinations with equivocal findings were excluded because of their inability to serve as a reference standard for CT or MRI findings. Finally, 112 DSA examinations that showed pseudoaneurysm or contrast extravasation were reviewed. The descriptive statistical results for the patients whose examinations were evaluated are shown in Table 1. Interrater Reliability The kappa value for the evaluation of pseudoaneurysm was 0.861, whereas that for aircontaining necrotic tissue was 0.801, that for TABLE 2: Hemorrhage Sites Identified by Digital Subtraction Angiography and the Consensus on CT or MRI Prediction of Respective Primary Cancers and Arteries Hemorrhage Sites Primary Cancer by Location Nasopharynx Oral Cavity Oropharynx Hypopharynx All Consensus ICA (73.1) Proximal ECA (33.3) Downstream ECA (86.8) CCA (69.6) Other (0) All Consensus 21 (77.8) 39 (78.0) 11 (84.6) 12 (54.5) Note Except where otherwise indicated, data are no. of bleeding events. ICA = internal carotid artery, ECA = external carotid artery, CCA = common carotid artery. 832 AJR:206, April 2016

5 Antecedent CT or MRI for Prediction of Hemorrhage Sites TABLE 3: Results of Prediction of Hemorrhage Sites CT or MRI Sign residual tumor was 0.797, and that for prediction of hemorrhage sites was These results are considered excellent for the individual CT or MRI signs and for prediction of the hemorrhage site by two radiologists. Relationship Between Primary Cancers, Bleeding Frequency, and Hemorrhage Sites Data on the primary cancers noted, the numbers of patients evaluated, and the respective hemorrhage sites are presented in Tables 1 and 2. Twenty-two patients underwent two or three DSA examinations for which results were positive, for a total of 48 examinations. Bleeding events noted in patients with nasopharyngeal cancers occurred more often in the ICA or the downstream ECA from the internal maxillary artery. For patients with oropharyngeal, hypopharyngeal, and oral cavity cancers, bleeding events occurred more often in the carotid system ipsilateral to the primary cancer, and they sometimes occurred from the contralateral downstream ECA. Predictive Ability The hemorrhage sites shown on DSA are listed in Table 2. The ECA was divided into the proximal ECA and the downstream ECA, whereas other arteries indicated bleeding from the vertebral artery, the thyrocervical trunk, or the unanticipated contralateral downstream ECA. Data indicated that the No Consensus No. (%) of Bleeding Events Consensus No sign (n = 6) 6 (100) 0 (0) Only one sign (n = 42) 10 (23.8) 32 (76.2) Any two signs (n = 54) 13 (24.1) 41 (75.9) All three signs (n = 10) 0 (0) 10 (100) All (n = 112) 29 (25.9) 83 (74.1) consensus for predicting hemorrhage sites on the basis of CT or MRI findings versus DSA results was better for the downstream ECA (86.8%), the ICA (73.1%), and the CCA (69.6%) than for the proximal ECA (33.3%) and other vessels (0%). For the four bleeding events that occurred in the proximal ECA, the last performed CT examinations were performed in a long time frame ( days) before DSA in three events. For one case of hypopharyngeal cancer, a CT study performed 14 days before DSA was performed showed similar tissue necrosis severity around the CCA and the proximal ECA; in this case, the CCA was chosen by both radiologists as the hemorrhage site. None of the four bleeding events associated with other bleeding arteries, including the contralateral downstream ECA and the thyrocervical artery, was predicted correctly. The consensus on the hemorrhage site between the last CT or MRI study performed and DSA performed for evaluation of primary cancers is listed in Table 2. No difference was observed among the primary cancers and the consensus values for prediction by CT or MRI and DSA results (p = 0.126, by Fisher exact test). However, a statistically significant difference was detected between hypopharyngeal and nonhypopharyngeal cancers (p = 0.019, by chisquare test). Of the 112 events for which DSA results were positive, contrast-enhanced CT, CTA, and MRI examinations were performed for 70, 17, and 25 events, respectively. The consensus values for overall CT and MRI were 75.9% and 68%, respectively, with no statistically significant difference noted (p = 0.429). The consensus values for CTA and contrast-enhanced CT were 92.9% and 72.6%, respectively, with no statistically significant difference noted (p = 0.172). The consensus value for CTA or MRI was 94.1%, and that for contrast-enhanced CT or MRI was 70.5%, with no statistically significant difference noted (p = 0.067). Imaging Signs and Optimal Timing of CT or MRI Examination for Better Prediction of Hemorrhage Sites According to the Mantel-Haenszel test of linear association, the more positive signs that were present, the more accurate was the consensus between CT or MRI and DSA (Table 3). This finding was considered statistically significant (p = 0.002). On the basis of the sensitivity and specificity values obtained for the respective bleeding arteries, we used the consensus on the interval between the time that CT or MRI was performed and the time that and DSA was performed to generate an ROC curve. The AUC was 0.738, which indicated the mean correct prediction of the number of days denoting the interval by consensus. Furthermore, by using the Youden index, we determined that 30 days was the optimal cutoff point for agreement on prediction of the hemorrhage site by CT or MRI. Details regarding the sensitivity and specificity of CT or MRI for prediction of the hemorrhage site from the respective arteries are shown in Table 4. The sensitivity of CT or MRI for the prediction of bleeding events in the respective arteries was found to be statistically significantly different for the ICA (p = 0.001) and the CCA (p = 0.007), when the interval was less than 30 days versus more than 30 days (Fig. 1). TABLE 4: Sensitivity and Specificity of CT or MRI for the Prediction of Bleeding Events in Respective Arteries Artery Overall Sensitivity (%) Overall Specificity (%) Interval Cutoff Sensitivity (%) Specificity (%) 30 Days >30 Days 30 Days >30 Days Sensitivity Specificity ICA NSS Proximal ECA NSS NSS Downstream ECA NSS NSS CCA NSS Note ICA = internal carotid artery, NSS = not statistically significant, ECA = external carotid artery, CCA = common carotid artery. p AJR:206, April

6 Ku et al. Relationship Between Equivocal Results of Digital Subtraction Angiography and the Interval Between the Last CT or MRI Performed and a Subsequently Performed Digital Subtraction Angiography With Positive Results Of a total of 123 DSA examinations, 11 examinations (8.9%) that were performed for 10 patients revealed equivocal findings and that were followed within 3 days by another bleeding event for which DSA findings were positive. The CT or MRI findings and the positive results of DSA examinations subsequently performed for seven patients (eight DSA examinations had equivocal results) had good consensus regarding hemorrhage sites, with all CT and MRI studies having been performed less than 30 days before the bleeding event occurred. On the other hand, three patients for whom there was no consensus on prediction of hemorrhage sites had an interval of more than 30 days between the time the last CT or MRI examination was performed and the time that a DSA examination was performed. Discussion After treatment of head and neck cancers, local inflammation and residual tumors are common findings on surveillance CT and MRI studies, and both findings tend to cause tissue necrosis and infection of varying extents [3, 8]. Most of these adverse effects are controllable, but some will progress to overt bleeding events. The frequency of bleeding is not high; however, when bleeding occurs, it requires urgent treatment because of the grave morbidity and mortality that result [9, A 10]. Given that most CT and MRI studies are not performed at the same time that DSA examinations are conducted, the findings can change during the interval between CT or MRI and DSA examination. The hypothesis of the present study was that a timely antecedent CT or MRI study could help predict the hemorrhage site in patients with posttreatment head and neck cancers. Our results suggest that CT or MRI studies, if performed within 30 days of the time that a bleeding event occurs, could be useful in the prediction of hemorrhage sites from the CCA and the ICA (Fig. 1). Prediction of hemorrhage sites by CT or MRI is good, except in the case of hypopharyngeal cancers. For bleeding events resulting from hypopharyngeal cancers, the correct prediction rate was only approximately 60%. This presumably resulted from hypopharyngeal cancers being inherently located in the middle lateral neck and at similar distances from the CCA, the ICA, the proximal ECA, and even the downstream ECA, such as the superior thyroid artery. On CT or MR images that show large areas of necrotic tissue, different carotid arteries will likely be affected simultaneously (Fig. 2). As mentioned in the Materials and Methods section of the present study, we prioritized choosing the artery considered to be at greater risk for brain infarction, such as the CCA or the ICA, instead of the downstream ECA, when the probability of bleeding occurring was similar. It is conceivable that the rate of correctly predicting hemorrhage sites was lower when bleeding events B C Fig year-old man with posttreatment hypopharyngeal cancer with two bleeding events in right carotid system. A C, CT scan acquired 3 months before first bleeding event shows only local treatment-related change in right pyriform sinus away from right common carotid artery (CCA) (arrow, A). First bleeding event from superior thyroid artery is considered complication of local treatment (arrow, B), and CT finding shown in panel A was unable to predict second bleeding event from CCA sidewall (arrow, C) that occurred 12 days later. recurred in different arteries, when predictions were based on findings from the same last CT or MRI study performed. For nonhypopharyngeal cancers, a prediction rate of approximately 80% still seems low. Because, on the basis of our findings, we are aware of the benefit of performing crosssectional CT or MRI studies before bleeding events occur, at our institution we are performing more contrast-enhanced CT or CTA studies within 1 day of the DSA examination performed at the time of hemorrhage. Recently, this did indeed increase the prediction rate. However, performing an immediate CT examination is not always possible or suitable for patients with bleeding and unstable vital signs [4]. Further study of the timing of the last CT or MRI examination performed is still necessary. In addition, DSA findings can often be equivocal, such as luminal narrowing or increased vascular blush, although we also know that arterial narrowing or even occlusion can be seen in cases of impending carotid blowout with hemorrhage into the vessel wall. Some equivocal findings noted on the initial DSA examination indeed eventually were identified as hemorrhagic sites on subsequent DSA examinations. However, we consider equivocal DSA findings should be regarded as clues only because they are insufficient to serve as strong evidence of hemorrhagic sites for treatment. The uncertainty of decision making (i.e., preemptive stenting of the CCA or the ICA) in the angiography suite reminds us to review the prior CT and MRI findings to explain the bleeding 834 AJR:206, April 2016

7 Antecedent CT or MRI for Prediction of Hemorrhage Sites episode. On the basis of our results, prediction of hemorrhagic sites by timely antecedent CT or MRI can facilitate more confident use of the interventional procedure. On the other hand, although our results showed that enhanced CT, CTA, and MRI were not statistically significantly different in predicting hemorrhage sites, we probably cannot overstate this result because the group numbers are not similar. However, we believe that the advantages of CTA, including thin slice section and better arterial details, such as identifying bovine aortic arches, will facilitate the interventional procedure. If no CT or MRI study performed within 30 days of the bleeding event is available, we suggest performing emergent CTA. We intentionally divided the ECA into proximal and distal portions because of the differences in endovascular treatment. However, it was revealed that correct prediction of the hemorrhage site was better only for the ICA, downstream ECA, and CCA, and not for the proximal ECA and other arteries. The suboptimal consensus on proximal ECA bleeding may result from the smaller number of bleeding events and the longer interval between the time that the CT or MRI study was performed and the time that the DSA was performed, leading to biased results. The fact that none of the four bleeding events that occurred in other arteries was correctly predicted by CT or MRI is considered reasonable (Fig. S1, a supplemental image, can be viewed in the AJR electronic supplement to this article, available at It is a basic clinical concept that the hemorrhage site is most commonly associated with the location of the primary cancer because of the initially aggressive management used. For example, bleeding occurring from the ICA and the internal maxillary artery in the skull base occurred exclusively in patients with nasopharyngeal cancers, whereas hemorrhage occurring from the downstream ECA in the floor of the mouth was more likely in patients with oral cavity cancers [11] (Fig. S2, a supplemental image, can be viewed in the AJR electronic supplement to this article, available at org). Because we can readily obtain patients clinical history of primary cancers when reviewing CT or MRI examinations, predicting hemorrhage sites on the basis of this approach might be practically useful. When recurrent cancers involve different levels (e.g., from the oropharynx to the hypopharynx) or are associated with secondary primary cancers, diagnosis of the bleeding sites will be challenging. These recurrent or secondary primary cancers presenting years later might be distant from the primary cancers. Our results suggest that the most recent CT or MRI examination should be evaluated; the updated CT or MRI findings will offer more accurate prediction of the hemorrhage site (Fig. 3 and Fig. S3; the latter figure is a supplemental image and can be viewed in the AJR electronic supplement to this article, available at org). In this setting, in cases when patients do not have a CT or MRI study performed within 30 days of the bleeding event, emergent CTA should be performed, if possible, before DSA is performed. We did not assess the correlation of CT or MRI findings with clinical and angiographic adverse events [3, 12]. The clinical scenario was always uncontrollable oral or neck bleeding, and the interventionists were consulted to cease the bleeding. Almost every patient with air-containing necrotic tissue will eventually experience a bleeding episode, unless percutaneous drainage or surgical débridement is performed in time. In our daily practice, tissue necrosis or pseudoaneurysm identified on CT or MRI prompts more aggressive treatment, such as percutaneous drainage, surgical débridement, or endovascular intervention, before actual carotid blowout occurs. The present study has several limitations. First, there was no control group of patients with head and neck cancers without active bleeding to highlight the character of CT or MRI findings in bleeding cases. Second, although our results showed that the more positive signs that were present, the more accurate was the consensus between CT or MRI and DSA, we could not determine which of the three CT or MRI signs contributed most to the consensus between findings from CT or MRI and DSA. Third, the presence of residual tumor was not verified by pathologic findings. Fourth, we chose three CT and MRI abnormalities related to bleeding (pseudoaneurysm, air-containing necrotic tissue, and residual tumor), in an attempt to predict hemorrhage sites. It is generally agreed that most patients have pseudoaneurysms or carotid blowout develop as a result of salivary secretions eroding the vessel wall, infection from open communication with the skin surface, or tumor eroding the vessel wall. When a pseudoaneurysm is the cause of a bleeding event, there is either tissue necrosis or residual tumor. Our results did not prove the cause, and this may merit further exploration and analysis. Finally, we assumed that tissue necrosis and residual tumor would not only be useful for treatment planning when a bleeding event occurs but would also indicate the need for preemptive surgical or endoscopic débridement and systemic antibiotic treatments when CT or MRI findings are available before clinical symptoms develop [8]. This is not verified by our results, but it is planned as a topic for future study. Conclusion The imaging findings from timely antecedent CT or MRI studies are useful in predicting hemorrhage sites in patients with posttreatment head and neck cancers complicated by bleeding events. The consensus was higher for patients with oral cavity, nasopharyngeal, and oropharyngeal cancers, but consensus was mediocre for patients with hypopharyngeal cancers. For all patients, the last CT or MRI examinations performed before DSA are optimal for predicting hemorrhage sites and complementing equivocal DSA results if they are performed within 30 days of the bleeding events. If no CT or MRI study performed within 30 days of the bleeding event is available, we suggest performing emergent CTA to obtain thin slice sections and better arterial details. References 1. Zhao LB, Shi HB, Park S, et al. Acute bleeding in the head and neck: angiographic findings and endovascular management. AJNR 2014; 35: Mazumdar A, Derdeyn CP, Holloway W, Moran CJ, Cross DT 3rd. Update on endovascular management of the carotid blowout syndrome. Neuroimaging Clin N Am 2009; 19: Chang FC, Luo CB, Lirng JF, et al. Evaluation of the outcomes of endovascular management for patients with head and neck cancers and associated carotid blowout syndrome of the external carotid artery. Clin Radiol 2013; 68:e561 e Lee CW, Yang CY, Chen YF, Huang A, Wang YH, Liu HM. CT angiography findings in carotid blowout syndrome and its role as a predictor of 1-year survival. AJNR 2014; 35: Fajardo LF, Lee A. Rupture of major vessels after radiation. Cancer 1975; 36: Chin SC, Jen YM, Chen CY, Som PM. Necrotic nasopharyngeal mucosa: an ominous MR sign of a carotid artery pseudoaneurysm. AJNR 2005; AJR:206, April

8 Ku et al. 26: Biffl WL, Moore EE, Offner PJ, Brega KE, Franciose RJ, Burch JM. Blunt carotid arterial injuries: implications of a new grading scale. J Trauma 1999; 47: Cavel O, Fliss DM, Segev Y, Zik D, Khafif A, Landsberg R. The role of the otorhinolaryngologist in the management of central skull base osteomyelitis. Am J Rhinol 2007; 21: McDonald MW, Moore MG, Johnstone PA. Risk of carotid blowout after reirradiation of the head and neck: a systematic review. Int J Radiat Oncol Biol Phys 2012; 82: Chen YJ, Wang CP, Wang CC, Jiang RS, Lin JC, Liu SA. Carotid blowout in patients with head and neck cancer: associated factors and treatment outcomes. Head Neck 2015; 37: Luo CB, Teng MM, Chang FC. Radiation acute carotid blowout syndromes of the ascending pharyngeal and internal carotid arteries in nasopharyngeal carcinoma. Eur Arch Otorhinolaryngol 2006; 263: Chaloupka JC, Putman CM, Citardi MJ, Ross DA, Sasaki CT. Endovascular therapy for the carotid blowout syndrome in head and neck surgical patients: diagnostic and managerial considerations. AJNR 1996; 17: FOR YOUR INFORMATION A data supplement for this article can be viewed in the online version of the article at: AJR:206, April 2016