- Email: [email protected]
Optimizing screening for blunt cerebrovascular injuries
Optimizing Screening for Blunt Cerebrovascular Injuries Walter L. Biffl, MD, Ernest E. Moore, MD, Patrick J. Offner, MD, Kerry E. Brega, MD, Reginald J. Franciose, MD, J. Paul Elliott, MD, Jon M. Burch, MD, Denver, Colorado
BACKGROUND: The recognition that early diagnosis and intervention, prior to ischemic neurologic injury, has the potential to improve outcome following blunt cerebrovascular injuries (BCVI), led to a policy of aggressive screening for these injuries. The resultant epidemic of BCVI has created a dilemma, as widespread screening is impractical. We sought to identify independent predictors of BCVI, to focus resources. METHODS: Cerebral arteriography was performed based on signs or symptoms of BCVI, or in asymptomatic patients with high-risk mechanisms (hyperextension, hyperflexion, direct blow) or injury patterns. Logistic regression analysis identified independent predictors. RESULTS: A total of 249 patients underwent arteriography; 85 (34%) had injuries. Independent predictors of carotid arterial injury were Glasgow coma score ≤6, petrous bone fracture, diffuse axonal brain injury, and LeFort II or III fracture. Having one of these factors in the setting of a high-risk mechanism was associated with 41% risk of injury. Of patients with cervical spine fracture, 39% had vertebral arterial injury. CONCLUSIONS: Patients sustaining high-risk injury mechanisms or patterns should be screened for BCVI. In the face of limited resources, screening efforts should be focused on those with high-risk predictors. Am J Surg. 1999;178:517–522. © 1999 by Excerpta Medica, Inc.
B
lunt cerebrovascular injuries (BCVI) have the potential for devastating complications. Historically, the vast majority have been diagnosed only after the appearance of neurologic deficits.1– 8 Based on a multicenter review of blunt carotid injuries (BCI),6 we proposed that these injuries were often unrecognized or masked by brain injury. To test this hypothesis, we began a prospective analysis of screening at our center in November 1994.9 In this phase, all patients undergoing arteriography to rule out traumatic aortic injuries were additionally evaluated
From the Departments of Surgery (WLB, EEM, PJO, RJF, JMB) and Neurosurgery (KEB, JPE), Denver Health Medical Center and University of Colorado Health Sciences Center, Denver, Colorado. Requests for reprints should be addressed to Walter L. Biffl, MD, Department of Surgery, Box 0206, Denver Health Medical Center, 777 Bannock Street, Denver, Colorado 80204-4507. Presented at the 51st Annual Meeting of the Southwestern Surgical Congress, Coronado, California, April 18 –21, 1999.
© 1999 by Excerpta Medica, Inc. All rights reserved.
with selective cerebral arteriography. The incidence of BCI in this select group was 3.5%; notably, there was no clinical suspicion of BCI prior to angiography in 3 of the 6 injured patients. Recognizing further that the majority of BCI have a latent period prior to the appearance of clinical symptoms,2,3,8,10,11 and with evidence from Memphis8 that neurologic outcome could be improved with anticoagulation, a more aggressive policy of screening was subsequently adopted at our center. Indeed, this policy has uncovered an epidemic of BCI.12 Unfortunately, a clinical dilemma has emerged. Although our data12 suggest that early diagnosis and therapy— before the appearance of neurologic deficits— contribute to improved neurologic outcome following BCI, a liberal screening approach is logistically impractical in most centers. Moreover, currently only cerebral arteriography and magnetic resonance angiography (MRA) are reliable screening tests.13 Thus, in remote facilities, patients may need to be transferred to other institutions to exclude BCVI. Our screening criteria, based on anatomic and mechanistic considerations, might be considered too inclusive. We hypothesized that a subgroup at high risk for BCVI could be identified, who warrant screening evaluation. Specifically, the purpose of this study was to determine the association between various injury mechanisms and patterns and BCVI.
METHODS Patients Denver Health Medical Center is a certified urban level I trauma center with pediatric commitment, and serves as the Rocky Mountain regional trauma center for Colorado and adjoining regions. The number of trauma admissions during the study period (January 1990 through September 1998) has been stable at 3000 to 3200 patients per year, and 86% of admissions have resulted from blunt injury. Our trauma registry records patients at the time of their hospitalization, and was employed to identify patients screened and diagnosed with BCVI prior to August 1996. Subsequently, patients undergoing cerebral arteriography to exclude BCVI have been identified, and specific data collected prospectively. Diagnosis The diagnosis of BCVI is confirmed by four-vessel cerebral arteriography in all cases. Digital subtraction techniques are used, and all studies include the aortic arch and cerebral vessel origins. Trauma patients undergo emergent arteriography if any of the following signs or symptoms suggestive of cerebrovascular injury are present: (a) hemorrhage—from mouth, nose, ears or wounds— of potential arterial origin; (b) expanding cervical hematoma; (c) cer0002-9610/99/$–see front matter PII S0002-9610(99)00245-7
517
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
RESULTS
TABLE I Screening Criteria for Blunt Cerebrovascular Injury Injury mechanism ● Severe cervical hyperextension/rotation or hyperflexion, particularly if associated with Displaced midface or complex mandibular fracture Closed head injury consistent with diffuse axonal injury ● Near-hanging resulting in anoxic brain injury Physical signs ● Seat belt abrasion or other soft tissue injury of the anterior neck resulting in significant swelling or altered mental status Fracture in proximity to internal carotid or vertebral artery ● Basilar skull fracture involving the carotid canal ● Cervical vertebral body fracture
vical bruit in a patient ⬍50 years old; (d) evidence of cerebral infarction on CT scan; or (e) unexplained or incongruous central or lateralizing neurologic deficit, transient ischemic attack (TIA), amaurosis fugax, or Horner’s syndrome. In August 1996 we began to screen at-risk asymptomatic patients (ie, no suggestive signs or symptoms) for BCVI. The criteria for screening arteriography are listed in Table I. Follow-up arteriography is performed within 7 to 10 days when possible, to evaluate efficacy of the initial therapy. Data Analysis All patients undergoing cerebral arteriography to exclude BCVI were studied. Demographic information, injury mechanisms and associated injuries were analyzed to identify risk factors for BCVI. Statistical analysis was performed on an IBM compatible personal computer using StatMost 32 for Windows 95 (DataMost Corp., Sandy, Utah) and SPSS 9.0 for Windows (SPSS, Inc., Chicago, Illinois). Means of continuous data were compared using Student’s t test. Categorical data were compared using Fisher’s exact test or the chi-square test, where appropriate. Risk factors for BCI or blunt vertebral artery injuries (BVI) were evaluated in univariate logistic regression analyses. Craniocervical risk factors that were associated with BCI or BVI with a P value ⱕ0.20 were entered into multiple logistic regression analysis. This cutoff was chosen to exclude variables of questionable importance, but to include variables of potential clinical relevance. In the multiple logistic regression analysis, significance was evaluated at the 0.05 level. Adjusted odds ratios and 95% confidence intervals were calculated for each of the independent predictors. The conditional probability of having a BCI or BVI was calculated as follows: if four risk factors were identified (X1, X2, X3, X4), then five coefficients (a, B1, B2, B3, B4) were derived from the analysis, and the conditional probability (CP) of having BCVI is:
Patients From January 1990 through September 1998, cerebral arteriography was performed in 249 patients to exclude BCVI; 85 (34%) were diagnosed with injuries. Sixty-five patients had carotid injuries, 10 had vertebral injuries, and 10 had both carotid and vertebral injuries. Carotid injuries were bilateral in 32 patients, and vertebral injuries bilateral in 5. Forty patients initially presented with signs or symptoms of BCVI; 28 (70%) had injuries. Among 209 asymptomatic patients, we diagnosed injuries in 57 (27%). The mean age of the 249 total patients was 37.9 ⫾ 1.1 years (range 5 to 83). Those with BCVI were younger (35.3 ⫾ 1.6 years) than those without BCVI (39.2 ⫾ 1.4; P ⬍0.05). Males comprised 79% of the total group (196 patients). Sixty-four of 196 (33%) males had injuries, compared with 21 of 53 (40%) females (P ⬎0.05). Mechanism of injury was motor vehicle crash in 113 (45%), pedestrian struck in 30 (12%), motorcycle crash in 28 (11%), fall in 25 (10%), assault in 15 (6%), skier versus tree in 12 (5%), and other mechanisms (bicycle crash, near-hanging, strangulation, construction accidents, and miscellaneous) in 26 (10%) patients. There was no difference in mechanism between the groups with and without injuries. Of those involved in motor vehicle crashes, 25% of the BCVI group were using passive restraints, compared with 29% of the group without BCVI (P ⬎0.05). Risk Factors Risk factors for BCVI are listed in Table II, which separately compares the groups with and without BCI and BVI, and lists the P value for each factor according to univariate analysis. In this analysis, significant risk factors for BCI were younger age, lower Glasgow Coma Score (GCS), GCS ⱕ6, diffuse axonal brain injury (DAI), presence of any head injury, petrous bone fracture, LeFort II or III fracture, associated chest injury, and associated abdominal injury. Significant risk factors for BVI were cervical spine fracture, any spine fracture, and associated abdominal injury. Multiple logistic regression analysis included all categorical variables with a P value ⱕ0.20 by univariate analysis. However, because of their questionable clinical relevance, the following factors were excluded despite low P values: associated chest injury, associated abdominal injury, and “any head injury.” Independent predictors of BCI by this analysis were GCS ⱕ6, petrous bone fracture, DAI, and LeFort II or III fracture (Table III). The only independent predictor of BVI was cervical spine fracture. Table IV lists the conditional probability of BCI and BVI in the presence of various risk factors; probability ranges reflect individual risk factors and combinations thereof.
COMMENTS CP ⫽ 1/(1 ⫹ e⫺z), where z ⫽ a ⫹ B1(X1) ⫹ B2 (X2) ⫹ B3(X3) ⫹ B4(X4) The risk factors X1, X2, X3, and X4 were assigned a value of “1” if present and “0” if absent. Continuous data are expressed as mean ⫾ the standard error of the mean. 518
Blunt cerebrovascular injuries (BCVI) have the potential for devastating complications. Early reports collectively established mortality rates of BCI to be 28%, with 58% of survivors suffering severe neurologic sequelae.1–3 Subsequent multicenter reviews corroborated these disconcerting morbidity and mortality figures, and identified the incidence of BCI to be 0.08% to 0.17% among patients
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
TABLE II Results of Univariate Logistic Regression Analyses of Risk Factors for Blunt Carotid (BCI) and Vertebral Artery (BVI) Injury Factor Mean age Mean GCS GCS ⱕ6 Diffuse axonal injury Subdural hematoma Epidural hematoma Subarachnoid hemorrhage Brain contusion Skull fracture Any head injury Sphenoid fracture Petrous fracture Any basilar skull fracture Nasal fracture Mandible fracture Any midface fracture Mandible plus any midface fracture LeFort II or III fracture Tripod fracture Any facial fracture Cervical spine fracture Any spine fracture Chest injury Abdomen injury Pelvis fracture Extremity fracture
BCI (n ⴝ 75)
No BCI (n ⴝ 174)
P Value
BVI (n ⴝ 20)
No BVI (n ⴝ 229)
P Value
34.6 ⫾ 1.7* 8.7 ⫾ 0.6* 36 (48%)* 11 (15%)* 12 (16%) 9 (12%) 23 (31%) 21 (28%) 11 (15%) 45 (60%)* 10 (13%) 14 (19%)* 20 (27%) 9 (12%) 10 (13%) 20 (27%) 9 (12%) 8 (11%)* 6 (8%) 25 (33%) 10 (13%) 20 (27%) 34 (45%)* 20 (27%)* 12 (16%) 22 (29%)
39.3 ⫾ 1.3 11.2 ⫾ 0.4 45 (26%) 8 (5%) 17 (10%) 15 (9%) 40 (23%) 46 (26%) 23 (13%) 73 (42%) 20 (11%) 13 (7%) 39 (22%) 25 (14%) 19 (11%) 48 (28%) 10 (6%) 5 (3%) 16 (9%) 61 (35%) 21 (12%) 40 (23%) 52 (30%) 22 (13%) 20 (11%) 50 (29%)
⬍0.001 ⬍0.001 0.001 0.009 0.164 0.409 0.203 0.799 0.760 0.010 0.683 0.012 0.470 0.618 0.587 0.881 0.095 0.018 0.761 0.793 0.508 0.500 0.020 0.008 0.332 0.924
39.0 ⫾ 3.1 10.8 ⫾ 1.2 4 (20%) 3 (15%) 2 (10%) 1 (5%) 6 (30%) 4 (20%) 1 (5%) 10 (50%) 1 (5%) 0 (0%) 2 (10%) 4 (20%) 1 (5%) 3 (15%) 1 (5%) 0 (0%) 0 (0%) 4 (20%) 12 (60%)* 13 (65%)* 9 (45%) 7 (35%)* 2 (10%) 7 (35%)
37.8 ⫾ 1.1 10.4 ⫾ 0.3 75 (33%) 16 (7%) 27 (12%) 23 (10%) 57 (25%) 63 (28%) 33 (14%) 108 (47%) 29 (13%) 27 (12%) 57 (25%) 30 (13%) 28 (12%) 65 (28%) 18 (8%) 13 (6%) 22 (10%) 82 (36%) 19 (8%) 46 (20%) 77 (34%) 35 (15%) 30 (13%) 65 (28%)
0.363 0.383 0.801 0.208 0.811 0.475 0.615 0.471 0.265 0.808 0.332 0.719 0.151 0.393 0.352 0.208 0.648 0.728 0.747 0.163 ⬍0.001 ⬍0.001 0.309 0.030 0.693 0.533
* P ⬍ 0.05. GCS ⫽ Glasgow Coma Score.
TABLE III Predictive Model for Blunt Cerebrovascular Injuries Based on Multiple Logistic Regression Analysis Predictor
Estimate
P Value
Carotid artery injury Constant ⫺1.39 ⫾ 0.20 — GCS ⱕ6 0.68 ⫾ 0.31 0.029 Petrous fracture 0.97 ⫾ 0.43 0.025 Diffuse axonal injury 1.13 ⫾ 0.52 0.030 LeFort II or III fracture 1.31 ⫾ 0.61 0.033 Vertebral artery injury Constant ⫺3.37 ⫾ 0.38 — Cervical spine fracture 2.67 ⫾ 0.51 ⬍0.001
Odds Ratio (95% CI) — 1.98 (1.07–3.65) 2.64 (1.13–6.19) 3.09 (1.12–8.57) 3.70 (1.12–12.29) — 14.50 (5.30–39.63)
CI ⫽ confidence interval; GCS ⫽ Glasgow Coma Score.
admitted to trauma centers following blunt injury.4 –7 Based on our involvement in one such review,6 we hypothesized that these injuries were often unrecognized or masked by brain injury. Consequently, we began a prospective analysis of screening for BCI in 1994.9 By imaging the cervical vessels in patients undergoing postinjury thoracic aortography, we found BCI in 3.5%. Only one half of these patients had the injury suspected prior to the diagnostic test. Around the time our pilot screening study9 was coming to
a close, Fabian and colleagues8 published a large singleinstitution series of BCI. Their reported incidence of BCI was 0.33%, nearly triple that of previous reviews, confirming that a higher index of suspicion would lead to the more frequent diagnosis of these injuries. Despite the high index of suspicion among the Memphis group, however, 93% of their patients already had neurologic symptoms at the time of diagnosis; furthermore, while they demonstrated improved neurologic outcome when patients were fully anticoagulated with heparin, the ultimate outcome included a mortality of 31% and neurologic morbidity of 37%. Given these data, we pursued the potential to further improve neurologic outcomes by identifying and treating BCVI prior to the occurrence of cerebral ischemia. In fact, the characteristic latent period between the time of injury and the occurrence of cerebral ischemia makes screening a tenable concept. Krajewski and Hertzer,2 in their literature review of 96 patients with BCI, noted that 58% of patients first manifested symptoms 10 or more hours after the injury, and 36% became symptomatic 24 hours or more postinjury. Perry and colleagues3 similarly reported that 23% of patients first had symptoms more than 24 hours after injury. Mokri and colleagues10 had 9 of 18 BCI patients develop symptoms more than 3 days postinjury, with one injury manifesting 14 years later! Fabian and colleagues11 in 1990 reported 33% of their patients had a neurologic change ⱖ12 hours postinjury. These data collectively formed the basis for an aggressive
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
519
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
TABLE IV Conditional Probability of Blunt Cerebrovascular Injuries Based on Multiple Logistic Regression Analysis Carotid artery injury No risk factors Any one risk factor Any two risk factors Any three risk factors All four risk factors Vertebral artery injury No cervical spine fracture Cervical spine fracture
⫽ 20% ⫽ 33% to 48% ⫽ 56% to 74% ⫽ 80% to 88% ⫽ 93% ⫽ 3% ⫽ 33%
Risk factors were GCS ⱕ6, petrous fracture, diffuse axonal brain injury, or LeFort II or III fracture. GCS ⫽ Glasgow Coma Score.
policy of screening at our institution, which led to the discovery of an epidemic of BCI.12 Unfortunately, the documentation of a relatively high incidence of asymptomatic BCI, as well as BVI, has created a clinical dilemma. Although our data12 suggest that early diagnosis and therapy— before the appearance of neurologic deficits—may contribute to improved neurologic outcome, we recognize that a liberal screening approach is impractical in most centers. A major reason for this is the absence of a reliable noninvasive screening test. Duplex ultrasonography, considered by many to be the modality of choice for imaging the carotid arteries, had just an 86% sensitivity for BCI in a multicenter review.6 Duplex requires significant stenosis to detect flow disturbances, and does not clearly image the arteries at or above the skull base. Computed tomographic angiography (CTA) is attractive in that most multisystem trauma patients have indications for CT scanning.14 However, our experience has shown that it has a sensitivity similar to duplex ultrasonography.15 To image the cerebral vessels in their entirety with a slice thickness and pitch adequate for sufficiently sensitive reconstruction is not practical; in addition, there is bony artifact in the carotid canal, obscuring injuries in that segment of the artery. A reliable alternative to arteriography appears to be MRA,13 but a major impediment to its widespread application is the lack of availability at many institutions, as well as incompatibility with many orthopedic fixation devices and ventilatory equipment. Until MRA has been rigorously evaluated, cerebral arteriography remains the predominant screening modality, and therein lies the dilemma. There is an understandable reluctance to subject patients to an expensive, labor-intensive, invasive examination for screening purposes alone. Many physicians believe that strokes as a result of BCVI are rare events, and therefore have difficulty accepting the concept of aggressive screening. Moreover, although our data12 are supportive, there are no prospective, controlled studies demonstrating an outcome benefit related to early diagnosis and intervention. The concept of screening is not unique to our institution, and has typically been based on injury patterns. Eachempati and colleagues16 reported 8 patients diagnosed with BCI in whom the diagnosis was pursued because of basilar skull fracture through the carotid canal. Rogers and colleagues14 similarly screened asymptomatic patients with 520
specific injury patterns, including basilar skull and mandibular fractures. The recognized association between cervical spine fractures and vertebral artery injuries has similarly prompted screening. Following mid-cervical spine fracture or subluxation—including locked or perched facet, facet destruction with instability, or fracture involving the foramen transversarium—Willis and colleagues17 demonstrated a 46% incidence of vertebral artery injuries. Woodring and colleagues18 performed arteriography in 8 patients in whom transverse process fractures of the cervical vertebrae extended into the foramen transversarium; 7 (88%) had vertebral artery injuries. Our screening criteria are based on the presumed pathophysiology of BCVI, and include injury mechanisms in addition to associated high-risk fracture patterns (Table I). The most common mechanism whereby the internal carotid artery (ICA) is injured is hyperextension and contralateral rotation of the head and neck. This may be explained by anatomical relationships unique to the upper cervical region. The lateral articular processes and pedicles of the upper three cervical vertebrae project more anteriorly than do those of the lower four cervical vertebrae. Thus, the overlying distal cervical ICA is prone to stretch injury during cervical hyperextension.19 Rotation at the atlantoaxial joint may result in anterior movement of the contralateral C1 lateral mass, further exacerbating the stretch injury. In instances of acute cervical hyperflexion, as seen in motor vehicle crashes or falls, the artery may be compressed between the mandible and vertebral prominences.19 Seat belt injuries, strangulation injuries, and near-hangings may injure the common or ICA by a “direct blow” mechanism. Our screening criteria have taken these factors into account, and considered associated injury patterns as indicators of force vectors. In our series, the incidence of arteriographically demonstrable BCVI was 70% in the presence of suggestive signs or symptoms of cerebrovascular injury. Based on our screening criteria, 27% of selected asymptomatic patients were diagnosed with BCVI. This is a relatively high yield, particularly for an injury with such potentially devastating consequences. However, we acknowledge the seeming inclusive nature of the screening criteria and the impracticality of implementing such a screening protocol at many institutions. Therefore, we sought to identify markers for high risk to help focus diagnostic resources. While the dangers of BCI have been well documented, the neurologic morbidity and mortality associated with BVI are less clear. We tend to treat BVI less aggressively than BCI, generally instituting anticoagulation only in the setting of highgrade, bilateral, or symptomatic lesions. Therefore, for purposes of generating predictive models, we separated the analyses of risk factors for BCI and BVI. Multiple logistic regression analysis identified four individual factors that are independently associated with the existence of BCI. It is important to keep in mind that these factors must be in the context of a mechanism of injury with cervical hyperextension/rotation, hyperflexion, or a direct blow. In this setting, a GCS ⱕ6, DAI, petrous bone fracture, or LeFort II or III fracture identify the patient at particularly high risk for BCI. Petrous bone fractures represent a risk to the ICA in the carotid canal. Diffuse axonal injury is thought to be
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
related to traumatic head rotation or abrupt deceleration, also important in the pathogenesis of BCI. LeFort fractures are associated with significant frontal impact, again consistent with hyperextension or hyperflexion mechanisms. The significance of a GCS ⱕ6 lies in the fact that BCI have probably been historically underrecognized because of the presence of severe brain injury, masking neurologic changes. The severe brain injury reflects a significant blow to the head, which can result in cervical hyperextension or hyperflexion. Regardless of symptoms, in the presence of any one of the four risk factors, the incidence of BCI was 41%. The conditional probability of BCI of 20% with no risk factors (Table IV) reflects the fact that 20% of our patients with BCI had no significant associated injuries.20 The only independent predictor for BVI we identified was the presence of a cervical spine fracture, present in 65% of patients with BVI. If a patient presented with cervical spine fracture, there was a 39% chance of finding a BVI on arteriography. Further experience will clarify the high-risk cervical fracture levels and types. An analysis of the sensitivity and specificity of each screening criterion was impossible for a number of reasons. First, symptoms that in retrospect were attributed to BCVI were sometimes ascribed to head injury at the time of presentation, confusing the issue of whether the patient was considered truly symptomatic; second, patients with severe associated injuries may have fit the screening criteria but did not undergo arteriography because of cardiopulmonary instability or early demise; third, there are some patients who underwent arteriography who did not clearly meet the criteria, and some who arguably met the criteria but were not studied. In sum, we have identified markers of high risk for BCVI in patients sustaining trauma with cervical hyperextension/ rotation, hyperflexion, or direct blow mechanisms. Given the number of patients we have treated who sustained BCVI as a result of “trivial” trauma, or who had no associated injuries, these high-risk markers will fail to identify some injuries. However, they represent a basis for more aggressive screening at facilities that lack the specific resources for widespread definitive diagnostic testing. The question of optimal criteria will ultimately require a multicenter collaborative effort. Perhaps most importantly, accumulating data will help to determine whether this approach will improve patient outcomes.
REFERENCES 1. Yamada S, Kindt GW, Youmans JR. Carotid artery occlusion due to nonpenetrating injury. J Trauma. 1967;7:333–342. 2. Krajewski LP, Hertzer NR. Blunt carotid artery trauma: report of
DISCUSSION David V. Feliciano, MD (Atlanta, Georgia): Drs. Biffl and Moore and colleagues have presented another in a series of studies documenting the incidence of and risk factors suggestive of a diagnosis of blunt injury to the carotid and vertebral arteries. They have taken 12 patients with a diagnosis of blunt cerebrovascular injury documented between 1990 and 1996 after screening based on overt vascular signs or neurologic symptoms and added these 12 patients to a series of 28 symptomatic patients and
two cases and review of the literature. Ann Surg. 1980;191:341– 346. 3. Perry MO, Snyder WH, Thal ER. Carotid artery injuries caused by blunt trauma. Ann Surg. 1980;192:74 –77. 4. Davis JW, Holbrook TL, Hoyt DB, et al. Blunt carotid artery dissection: incidence, associated injuries, screening, and treatment. J Trauma. 1990;30:1514 –1517. 5. Martin RF, Eldrup-Jorgensen J, Clark DE, Bredenberg CE. Blunt trauma to the carotid arteries. J Vasc Surg. 1991;14:789 –795. 6. Cogbill TH, Moore EE, Meissner M, et al. The spectrum of blunt injury to the carotid artery: a multicenter perspective. J Trauma. 1994;37:473– 479. 7. Ramadan F, Rutledge R, Oller D, et al. Carotid artery trauma: a review of contemporary trauma center experiences. J Vasc Surg. 1995;21:46 –56. 8. Fabian TC, Patton JH Jr, Croce MA, et al. Blunt carotid injury: importance of early diagnosis and anticoagulant therapy. Ann Surg. 1996;223:513–525. 9. Prall JA, Brega KE, Coldwell DM, Breeze RE. Incidence of unsuspected blunt carotid artery injury. Neurosurgery. 1998;42:495– 499. 10. Mokri B, Piepgras DG, Houser OW. Traumatic dissections of the extracranial internal carotid artery. J Neurosurg. 1988;68:189 – 197. 11. Fabian TC, George SM Jr, Croce MA, et al. Carotid artery trauma: management based on mechanism of injury. J Trauma. 1990;30:953–963. 12. 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. 13. Bok APL, Peter JC. Carotid and vertebral artery occlusion after blunt cervical injury: the role of MR angiography in early diagnosis. J Trauma. 1996;40:968 –972. 14. Rogers FB, Baker EF, Osler TM, et al. Computed tomographic angiography as a screening modality for blunt cervical arterial injuries: preliminary results. J Trauma. 1999;46:380 –385. 15. Biffl WL, Moore EE, Mestek M. Computed tomographic angiography as a screening modality for blunt cervical arterial injuries: a cautionary note. J Trauma. 1999;47:438 – 439. Letter. 16. Eachempati SR, Vaslef SN, Sebastian MW, Reed RL II. Blunt vascular injuries of the head and neck: is heparinization necessary? J Trauma. 1998;45:997–1004. 17. Willis BK, Greiner F, Orrison WW, Benzel EC. The incidence of vertebral artery injury after midcervical spine fracture or subluxation. Neurosurgery. 1994;34:435– 442. 18. Woodring JH, Lee C, Duncan V. Transverse process fractures of the cervical vertebrae: are they insignificant? J Trauma. 1993; 34:797– 802. 19. Zelenock GB, Kazmers A, Whitehouse WM Jr, et al. Extracranial internal carotid artery dissections: noniatrogenic traumatic lesions. Arch Surg. 1982;117:425– 432. 20. Biffl WL, Moore EE, Offner PJ, et al. Blunt carotid arterial injuries: implications of a new grading scale. J Trauma. 1999;47: 845– 853.
209 asymptomatic patients subsequently undergoing arteriography based on the new criteria presented today since 1996. In summary, 85 patients over a period of 7 or 8 years, an extraordinary number, were found to have injury to the cerebrovascular vessels at one hospital. As stated, they then used univariate logistic regression analysis based on incidences, followed by multiple logistic regression analysis using the most significant categorical variables and univariate analysis. The authors concluded that independent
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
521
BACKGROUND: The recognition that early diagnosis and intervention, prior to ischemic neurologic injury, has the potential to improve outcome following blunt cerebrovascular injuries (BCVI), led to a policy of aggressive screening for these injuries. The resultant epidemic of BCVI has created a dilemma, as widespread screening is impractical. We sought to identify independent predictors of BCVI, to focus resources. METHODS: Cerebral arteriography was performed based on signs or symptoms of BCVI, or in asymptomatic patients with high-risk mechanisms (hyperextension, hyperflexion, direct blow) or injury patterns. Logistic regression analysis identified independent predictors. RESULTS: A total of 249 patients underwent arteriography; 85 (34%) had injuries. Independent predictors of carotid arterial injury were Glasgow coma score ≤6, petrous bone fracture, diffuse axonal brain injury, and LeFort II or III fracture. Having one of these factors in the setting of a high-risk mechanism was associated with 41% risk of injury. Of patients with cervical spine fracture, 39% had vertebral arterial injury. CONCLUSIONS: Patients sustaining high-risk injury mechanisms or patterns should be screened for BCVI. In the face of limited resources, screening efforts should be focused on those with high-risk predictors. Am J Surg. 1999;178:517–522. © 1999 by Excerpta Medica, Inc.
B
lunt cerebrovascular injuries (BCVI) have the potential for devastating complications. Historically, the vast majority have been diagnosed only after the appearance of neurologic deficits.1– 8 Based on a multicenter review of blunt carotid injuries (BCI),6 we proposed that these injuries were often unrecognized or masked by brain injury. To test this hypothesis, we began a prospective analysis of screening at our center in November 1994.9 In this phase, all patients undergoing arteriography to rule out traumatic aortic injuries were additionally evaluated
From the Departments of Surgery (WLB, EEM, PJO, RJF, JMB) and Neurosurgery (KEB, JPE), Denver Health Medical Center and University of Colorado Health Sciences Center, Denver, Colorado. Requests for reprints should be addressed to Walter L. Biffl, MD, Department of Surgery, Box 0206, Denver Health Medical Center, 777 Bannock Street, Denver, Colorado 80204-4507. Presented at the 51st Annual Meeting of the Southwestern Surgical Congress, Coronado, California, April 18 –21, 1999.
© 1999 by Excerpta Medica, Inc. All rights reserved.
with selective cerebral arteriography. The incidence of BCI in this select group was 3.5%; notably, there was no clinical suspicion of BCI prior to angiography in 3 of the 6 injured patients. Recognizing further that the majority of BCI have a latent period prior to the appearance of clinical symptoms,2,3,8,10,11 and with evidence from Memphis8 that neurologic outcome could be improved with anticoagulation, a more aggressive policy of screening was subsequently adopted at our center. Indeed, this policy has uncovered an epidemic of BCI.12 Unfortunately, a clinical dilemma has emerged. Although our data12 suggest that early diagnosis and therapy— before the appearance of neurologic deficits— contribute to improved neurologic outcome following BCI, a liberal screening approach is logistically impractical in most centers. Moreover, currently only cerebral arteriography and magnetic resonance angiography (MRA) are reliable screening tests.13 Thus, in remote facilities, patients may need to be transferred to other institutions to exclude BCVI. Our screening criteria, based on anatomic and mechanistic considerations, might be considered too inclusive. We hypothesized that a subgroup at high risk for BCVI could be identified, who warrant screening evaluation. Specifically, the purpose of this study was to determine the association between various injury mechanisms and patterns and BCVI.
METHODS Patients Denver Health Medical Center is a certified urban level I trauma center with pediatric commitment, and serves as the Rocky Mountain regional trauma center for Colorado and adjoining regions. The number of trauma admissions during the study period (January 1990 through September 1998) has been stable at 3000 to 3200 patients per year, and 86% of admissions have resulted from blunt injury. Our trauma registry records patients at the time of their hospitalization, and was employed to identify patients screened and diagnosed with BCVI prior to August 1996. Subsequently, patients undergoing cerebral arteriography to exclude BCVI have been identified, and specific data collected prospectively. Diagnosis The diagnosis of BCVI is confirmed by four-vessel cerebral arteriography in all cases. Digital subtraction techniques are used, and all studies include the aortic arch and cerebral vessel origins. Trauma patients undergo emergent arteriography if any of the following signs or symptoms suggestive of cerebrovascular injury are present: (a) hemorrhage—from mouth, nose, ears or wounds— of potential arterial origin; (b) expanding cervical hematoma; (c) cer0002-9610/99/$–see front matter PII S0002-9610(99)00245-7
517
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
RESULTS
TABLE I Screening Criteria for Blunt Cerebrovascular Injury Injury mechanism ● Severe cervical hyperextension/rotation or hyperflexion, particularly if associated with Displaced midface or complex mandibular fracture Closed head injury consistent with diffuse axonal injury ● Near-hanging resulting in anoxic brain injury Physical signs ● Seat belt abrasion or other soft tissue injury of the anterior neck resulting in significant swelling or altered mental status Fracture in proximity to internal carotid or vertebral artery ● Basilar skull fracture involving the carotid canal ● Cervical vertebral body fracture
vical bruit in a patient ⬍50 years old; (d) evidence of cerebral infarction on CT scan; or (e) unexplained or incongruous central or lateralizing neurologic deficit, transient ischemic attack (TIA), amaurosis fugax, or Horner’s syndrome. In August 1996 we began to screen at-risk asymptomatic patients (ie, no suggestive signs or symptoms) for BCVI. The criteria for screening arteriography are listed in Table I. Follow-up arteriography is performed within 7 to 10 days when possible, to evaluate efficacy of the initial therapy. Data Analysis All patients undergoing cerebral arteriography to exclude BCVI were studied. Demographic information, injury mechanisms and associated injuries were analyzed to identify risk factors for BCVI. Statistical analysis was performed on an IBM compatible personal computer using StatMost 32 for Windows 95 (DataMost Corp., Sandy, Utah) and SPSS 9.0 for Windows (SPSS, Inc., Chicago, Illinois). Means of continuous data were compared using Student’s t test. Categorical data were compared using Fisher’s exact test or the chi-square test, where appropriate. Risk factors for BCI or blunt vertebral artery injuries (BVI) were evaluated in univariate logistic regression analyses. Craniocervical risk factors that were associated with BCI or BVI with a P value ⱕ0.20 were entered into multiple logistic regression analysis. This cutoff was chosen to exclude variables of questionable importance, but to include variables of potential clinical relevance. In the multiple logistic regression analysis, significance was evaluated at the 0.05 level. Adjusted odds ratios and 95% confidence intervals were calculated for each of the independent predictors. The conditional probability of having a BCI or BVI was calculated as follows: if four risk factors were identified (X1, X2, X3, X4), then five coefficients (a, B1, B2, B3, B4) were derived from the analysis, and the conditional probability (CP) of having BCVI is:
Patients From January 1990 through September 1998, cerebral arteriography was performed in 249 patients to exclude BCVI; 85 (34%) were diagnosed with injuries. Sixty-five patients had carotid injuries, 10 had vertebral injuries, and 10 had both carotid and vertebral injuries. Carotid injuries were bilateral in 32 patients, and vertebral injuries bilateral in 5. Forty patients initially presented with signs or symptoms of BCVI; 28 (70%) had injuries. Among 209 asymptomatic patients, we diagnosed injuries in 57 (27%). The mean age of the 249 total patients was 37.9 ⫾ 1.1 years (range 5 to 83). Those with BCVI were younger (35.3 ⫾ 1.6 years) than those without BCVI (39.2 ⫾ 1.4; P ⬍0.05). Males comprised 79% of the total group (196 patients). Sixty-four of 196 (33%) males had injuries, compared with 21 of 53 (40%) females (P ⬎0.05). Mechanism of injury was motor vehicle crash in 113 (45%), pedestrian struck in 30 (12%), motorcycle crash in 28 (11%), fall in 25 (10%), assault in 15 (6%), skier versus tree in 12 (5%), and other mechanisms (bicycle crash, near-hanging, strangulation, construction accidents, and miscellaneous) in 26 (10%) patients. There was no difference in mechanism between the groups with and without injuries. Of those involved in motor vehicle crashes, 25% of the BCVI group were using passive restraints, compared with 29% of the group without BCVI (P ⬎0.05). Risk Factors Risk factors for BCVI are listed in Table II, which separately compares the groups with and without BCI and BVI, and lists the P value for each factor according to univariate analysis. In this analysis, significant risk factors for BCI were younger age, lower Glasgow Coma Score (GCS), GCS ⱕ6, diffuse axonal brain injury (DAI), presence of any head injury, petrous bone fracture, LeFort II or III fracture, associated chest injury, and associated abdominal injury. Significant risk factors for BVI were cervical spine fracture, any spine fracture, and associated abdominal injury. Multiple logistic regression analysis included all categorical variables with a P value ⱕ0.20 by univariate analysis. However, because of their questionable clinical relevance, the following factors were excluded despite low P values: associated chest injury, associated abdominal injury, and “any head injury.” Independent predictors of BCI by this analysis were GCS ⱕ6, petrous bone fracture, DAI, and LeFort II or III fracture (Table III). The only independent predictor of BVI was cervical spine fracture. Table IV lists the conditional probability of BCI and BVI in the presence of various risk factors; probability ranges reflect individual risk factors and combinations thereof.
COMMENTS CP ⫽ 1/(1 ⫹ e⫺z), where z ⫽ a ⫹ B1(X1) ⫹ B2 (X2) ⫹ B3(X3) ⫹ B4(X4) The risk factors X1, X2, X3, and X4 were assigned a value of “1” if present and “0” if absent. Continuous data are expressed as mean ⫾ the standard error of the mean. 518
Blunt cerebrovascular injuries (BCVI) have the potential for devastating complications. Early reports collectively established mortality rates of BCI to be 28%, with 58% of survivors suffering severe neurologic sequelae.1–3 Subsequent multicenter reviews corroborated these disconcerting morbidity and mortality figures, and identified the incidence of BCI to be 0.08% to 0.17% among patients
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
TABLE II Results of Univariate Logistic Regression Analyses of Risk Factors for Blunt Carotid (BCI) and Vertebral Artery (BVI) Injury Factor Mean age Mean GCS GCS ⱕ6 Diffuse axonal injury Subdural hematoma Epidural hematoma Subarachnoid hemorrhage Brain contusion Skull fracture Any head injury Sphenoid fracture Petrous fracture Any basilar skull fracture Nasal fracture Mandible fracture Any midface fracture Mandible plus any midface fracture LeFort II or III fracture Tripod fracture Any facial fracture Cervical spine fracture Any spine fracture Chest injury Abdomen injury Pelvis fracture Extremity fracture
BCI (n ⴝ 75)
No BCI (n ⴝ 174)
P Value
BVI (n ⴝ 20)
No BVI (n ⴝ 229)
P Value
34.6 ⫾ 1.7* 8.7 ⫾ 0.6* 36 (48%)* 11 (15%)* 12 (16%) 9 (12%) 23 (31%) 21 (28%) 11 (15%) 45 (60%)* 10 (13%) 14 (19%)* 20 (27%) 9 (12%) 10 (13%) 20 (27%) 9 (12%) 8 (11%)* 6 (8%) 25 (33%) 10 (13%) 20 (27%) 34 (45%)* 20 (27%)* 12 (16%) 22 (29%)
39.3 ⫾ 1.3 11.2 ⫾ 0.4 45 (26%) 8 (5%) 17 (10%) 15 (9%) 40 (23%) 46 (26%) 23 (13%) 73 (42%) 20 (11%) 13 (7%) 39 (22%) 25 (14%) 19 (11%) 48 (28%) 10 (6%) 5 (3%) 16 (9%) 61 (35%) 21 (12%) 40 (23%) 52 (30%) 22 (13%) 20 (11%) 50 (29%)
⬍0.001 ⬍0.001 0.001 0.009 0.164 0.409 0.203 0.799 0.760 0.010 0.683 0.012 0.470 0.618 0.587 0.881 0.095 0.018 0.761 0.793 0.508 0.500 0.020 0.008 0.332 0.924
39.0 ⫾ 3.1 10.8 ⫾ 1.2 4 (20%) 3 (15%) 2 (10%) 1 (5%) 6 (30%) 4 (20%) 1 (5%) 10 (50%) 1 (5%) 0 (0%) 2 (10%) 4 (20%) 1 (5%) 3 (15%) 1 (5%) 0 (0%) 0 (0%) 4 (20%) 12 (60%)* 13 (65%)* 9 (45%) 7 (35%)* 2 (10%) 7 (35%)
37.8 ⫾ 1.1 10.4 ⫾ 0.3 75 (33%) 16 (7%) 27 (12%) 23 (10%) 57 (25%) 63 (28%) 33 (14%) 108 (47%) 29 (13%) 27 (12%) 57 (25%) 30 (13%) 28 (12%) 65 (28%) 18 (8%) 13 (6%) 22 (10%) 82 (36%) 19 (8%) 46 (20%) 77 (34%) 35 (15%) 30 (13%) 65 (28%)
0.363 0.383 0.801 0.208 0.811 0.475 0.615 0.471 0.265 0.808 0.332 0.719 0.151 0.393 0.352 0.208 0.648 0.728 0.747 0.163 ⬍0.001 ⬍0.001 0.309 0.030 0.693 0.533
* P ⬍ 0.05. GCS ⫽ Glasgow Coma Score.
TABLE III Predictive Model for Blunt Cerebrovascular Injuries Based on Multiple Logistic Regression Analysis Predictor
Estimate
P Value
Carotid artery injury Constant ⫺1.39 ⫾ 0.20 — GCS ⱕ6 0.68 ⫾ 0.31 0.029 Petrous fracture 0.97 ⫾ 0.43 0.025 Diffuse axonal injury 1.13 ⫾ 0.52 0.030 LeFort II or III fracture 1.31 ⫾ 0.61 0.033 Vertebral artery injury Constant ⫺3.37 ⫾ 0.38 — Cervical spine fracture 2.67 ⫾ 0.51 ⬍0.001
Odds Ratio (95% CI) — 1.98 (1.07–3.65) 2.64 (1.13–6.19) 3.09 (1.12–8.57) 3.70 (1.12–12.29) — 14.50 (5.30–39.63)
CI ⫽ confidence interval; GCS ⫽ Glasgow Coma Score.
admitted to trauma centers following blunt injury.4 –7 Based on our involvement in one such review,6 we hypothesized that these injuries were often unrecognized or masked by brain injury. Consequently, we began a prospective analysis of screening for BCI in 1994.9 By imaging the cervical vessels in patients undergoing postinjury thoracic aortography, we found BCI in 3.5%. Only one half of these patients had the injury suspected prior to the diagnostic test. Around the time our pilot screening study9 was coming to
a close, Fabian and colleagues8 published a large singleinstitution series of BCI. Their reported incidence of BCI was 0.33%, nearly triple that of previous reviews, confirming that a higher index of suspicion would lead to the more frequent diagnosis of these injuries. Despite the high index of suspicion among the Memphis group, however, 93% of their patients already had neurologic symptoms at the time of diagnosis; furthermore, while they demonstrated improved neurologic outcome when patients were fully anticoagulated with heparin, the ultimate outcome included a mortality of 31% and neurologic morbidity of 37%. Given these data, we pursued the potential to further improve neurologic outcomes by identifying and treating BCVI prior to the occurrence of cerebral ischemia. In fact, the characteristic latent period between the time of injury and the occurrence of cerebral ischemia makes screening a tenable concept. Krajewski and Hertzer,2 in their literature review of 96 patients with BCI, noted that 58% of patients first manifested symptoms 10 or more hours after the injury, and 36% became symptomatic 24 hours or more postinjury. Perry and colleagues3 similarly reported that 23% of patients first had symptoms more than 24 hours after injury. Mokri and colleagues10 had 9 of 18 BCI patients develop symptoms more than 3 days postinjury, with one injury manifesting 14 years later! Fabian and colleagues11 in 1990 reported 33% of their patients had a neurologic change ⱖ12 hours postinjury. These data collectively formed the basis for an aggressive
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
519
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
TABLE IV Conditional Probability of Blunt Cerebrovascular Injuries Based on Multiple Logistic Regression Analysis Carotid artery injury No risk factors Any one risk factor Any two risk factors Any three risk factors All four risk factors Vertebral artery injury No cervical spine fracture Cervical spine fracture
⫽ 20% ⫽ 33% to 48% ⫽ 56% to 74% ⫽ 80% to 88% ⫽ 93% ⫽ 3% ⫽ 33%
Risk factors were GCS ⱕ6, petrous fracture, diffuse axonal brain injury, or LeFort II or III fracture. GCS ⫽ Glasgow Coma Score.
policy of screening at our institution, which led to the discovery of an epidemic of BCI.12 Unfortunately, the documentation of a relatively high incidence of asymptomatic BCI, as well as BVI, has created a clinical dilemma. Although our data12 suggest that early diagnosis and therapy— before the appearance of neurologic deficits—may contribute to improved neurologic outcome, we recognize that a liberal screening approach is impractical in most centers. A major reason for this is the absence of a reliable noninvasive screening test. Duplex ultrasonography, considered by many to be the modality of choice for imaging the carotid arteries, had just an 86% sensitivity for BCI in a multicenter review.6 Duplex requires significant stenosis to detect flow disturbances, and does not clearly image the arteries at or above the skull base. Computed tomographic angiography (CTA) is attractive in that most multisystem trauma patients have indications for CT scanning.14 However, our experience has shown that it has a sensitivity similar to duplex ultrasonography.15 To image the cerebral vessels in their entirety with a slice thickness and pitch adequate for sufficiently sensitive reconstruction is not practical; in addition, there is bony artifact in the carotid canal, obscuring injuries in that segment of the artery. A reliable alternative to arteriography appears to be MRA,13 but a major impediment to its widespread application is the lack of availability at many institutions, as well as incompatibility with many orthopedic fixation devices and ventilatory equipment. Until MRA has been rigorously evaluated, cerebral arteriography remains the predominant screening modality, and therein lies the dilemma. There is an understandable reluctance to subject patients to an expensive, labor-intensive, invasive examination for screening purposes alone. Many physicians believe that strokes as a result of BCVI are rare events, and therefore have difficulty accepting the concept of aggressive screening. Moreover, although our data12 are supportive, there are no prospective, controlled studies demonstrating an outcome benefit related to early diagnosis and intervention. The concept of screening is not unique to our institution, and has typically been based on injury patterns. Eachempati and colleagues16 reported 8 patients diagnosed with BCI in whom the diagnosis was pursued because of basilar skull fracture through the carotid canal. Rogers and colleagues14 similarly screened asymptomatic patients with 520
specific injury patterns, including basilar skull and mandibular fractures. The recognized association between cervical spine fractures and vertebral artery injuries has similarly prompted screening. Following mid-cervical spine fracture or subluxation—including locked or perched facet, facet destruction with instability, or fracture involving the foramen transversarium—Willis and colleagues17 demonstrated a 46% incidence of vertebral artery injuries. Woodring and colleagues18 performed arteriography in 8 patients in whom transverse process fractures of the cervical vertebrae extended into the foramen transversarium; 7 (88%) had vertebral artery injuries. Our screening criteria are based on the presumed pathophysiology of BCVI, and include injury mechanisms in addition to associated high-risk fracture patterns (Table I). The most common mechanism whereby the internal carotid artery (ICA) is injured is hyperextension and contralateral rotation of the head and neck. This may be explained by anatomical relationships unique to the upper cervical region. The lateral articular processes and pedicles of the upper three cervical vertebrae project more anteriorly than do those of the lower four cervical vertebrae. Thus, the overlying distal cervical ICA is prone to stretch injury during cervical hyperextension.19 Rotation at the atlantoaxial joint may result in anterior movement of the contralateral C1 lateral mass, further exacerbating the stretch injury. In instances of acute cervical hyperflexion, as seen in motor vehicle crashes or falls, the artery may be compressed between the mandible and vertebral prominences.19 Seat belt injuries, strangulation injuries, and near-hangings may injure the common or ICA by a “direct blow” mechanism. Our screening criteria have taken these factors into account, and considered associated injury patterns as indicators of force vectors. In our series, the incidence of arteriographically demonstrable BCVI was 70% in the presence of suggestive signs or symptoms of cerebrovascular injury. Based on our screening criteria, 27% of selected asymptomatic patients were diagnosed with BCVI. This is a relatively high yield, particularly for an injury with such potentially devastating consequences. However, we acknowledge the seeming inclusive nature of the screening criteria and the impracticality of implementing such a screening protocol at many institutions. Therefore, we sought to identify markers for high risk to help focus diagnostic resources. While the dangers of BCI have been well documented, the neurologic morbidity and mortality associated with BVI are less clear. We tend to treat BVI less aggressively than BCI, generally instituting anticoagulation only in the setting of highgrade, bilateral, or symptomatic lesions. Therefore, for purposes of generating predictive models, we separated the analyses of risk factors for BCI and BVI. Multiple logistic regression analysis identified four individual factors that are independently associated with the existence of BCI. It is important to keep in mind that these factors must be in the context of a mechanism of injury with cervical hyperextension/rotation, hyperflexion, or a direct blow. In this setting, a GCS ⱕ6, DAI, petrous bone fracture, or LeFort II or III fracture identify the patient at particularly high risk for BCI. Petrous bone fractures represent a risk to the ICA in the carotid canal. Diffuse axonal injury is thought to be
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
SCREENING FOR BLUNT CEREBROVASCULAR INJURIES/BIFFL ET AL
related to traumatic head rotation or abrupt deceleration, also important in the pathogenesis of BCI. LeFort fractures are associated with significant frontal impact, again consistent with hyperextension or hyperflexion mechanisms. The significance of a GCS ⱕ6 lies in the fact that BCI have probably been historically underrecognized because of the presence of severe brain injury, masking neurologic changes. The severe brain injury reflects a significant blow to the head, which can result in cervical hyperextension or hyperflexion. Regardless of symptoms, in the presence of any one of the four risk factors, the incidence of BCI was 41%. The conditional probability of BCI of 20% with no risk factors (Table IV) reflects the fact that 20% of our patients with BCI had no significant associated injuries.20 The only independent predictor for BVI we identified was the presence of a cervical spine fracture, present in 65% of patients with BVI. If a patient presented with cervical spine fracture, there was a 39% chance of finding a BVI on arteriography. Further experience will clarify the high-risk cervical fracture levels and types. An analysis of the sensitivity and specificity of each screening criterion was impossible for a number of reasons. First, symptoms that in retrospect were attributed to BCVI were sometimes ascribed to head injury at the time of presentation, confusing the issue of whether the patient was considered truly symptomatic; second, patients with severe associated injuries may have fit the screening criteria but did not undergo arteriography because of cardiopulmonary instability or early demise; third, there are some patients who underwent arteriography who did not clearly meet the criteria, and some who arguably met the criteria but were not studied. In sum, we have identified markers of high risk for BCVI in patients sustaining trauma with cervical hyperextension/ rotation, hyperflexion, or direct blow mechanisms. Given the number of patients we have treated who sustained BCVI as a result of “trivial” trauma, or who had no associated injuries, these high-risk markers will fail to identify some injuries. However, they represent a basis for more aggressive screening at facilities that lack the specific resources for widespread definitive diagnostic testing. The question of optimal criteria will ultimately require a multicenter collaborative effort. Perhaps most importantly, accumulating data will help to determine whether this approach will improve patient outcomes.
REFERENCES 1. Yamada S, Kindt GW, Youmans JR. Carotid artery occlusion due to nonpenetrating injury. J Trauma. 1967;7:333–342. 2. Krajewski LP, Hertzer NR. Blunt carotid artery trauma: report of
DISCUSSION David V. Feliciano, MD (Atlanta, Georgia): Drs. Biffl and Moore and colleagues have presented another in a series of studies documenting the incidence of and risk factors suggestive of a diagnosis of blunt injury to the carotid and vertebral arteries. They have taken 12 patients with a diagnosis of blunt cerebrovascular injury documented between 1990 and 1996 after screening based on overt vascular signs or neurologic symptoms and added these 12 patients to a series of 28 symptomatic patients and
two cases and review of the literature. Ann Surg. 1980;191:341– 346. 3. Perry MO, Snyder WH, Thal ER. Carotid artery injuries caused by blunt trauma. Ann Surg. 1980;192:74 –77. 4. Davis JW, Holbrook TL, Hoyt DB, et al. Blunt carotid artery dissection: incidence, associated injuries, screening, and treatment. J Trauma. 1990;30:1514 –1517. 5. Martin RF, Eldrup-Jorgensen J, Clark DE, Bredenberg CE. Blunt trauma to the carotid arteries. J Vasc Surg. 1991;14:789 –795. 6. Cogbill TH, Moore EE, Meissner M, et al. The spectrum of blunt injury to the carotid artery: a multicenter perspective. J Trauma. 1994;37:473– 479. 7. Ramadan F, Rutledge R, Oller D, et al. Carotid artery trauma: a review of contemporary trauma center experiences. J Vasc Surg. 1995;21:46 –56. 8. Fabian TC, Patton JH Jr, Croce MA, et al. Blunt carotid injury: importance of early diagnosis and anticoagulant therapy. Ann Surg. 1996;223:513–525. 9. Prall JA, Brega KE, Coldwell DM, Breeze RE. Incidence of unsuspected blunt carotid artery injury. Neurosurgery. 1998;42:495– 499. 10. Mokri B, Piepgras DG, Houser OW. Traumatic dissections of the extracranial internal carotid artery. J Neurosurg. 1988;68:189 – 197. 11. Fabian TC, George SM Jr, Croce MA, et al. Carotid artery trauma: management based on mechanism of injury. J Trauma. 1990;30:953–963. 12. 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. 13. Bok APL, Peter JC. Carotid and vertebral artery occlusion after blunt cervical injury: the role of MR angiography in early diagnosis. J Trauma. 1996;40:968 –972. 14. Rogers FB, Baker EF, Osler TM, et al. Computed tomographic angiography as a screening modality for blunt cervical arterial injuries: preliminary results. J Trauma. 1999;46:380 –385. 15. Biffl WL, Moore EE, Mestek M. Computed tomographic angiography as a screening modality for blunt cervical arterial injuries: a cautionary note. J Trauma. 1999;47:438 – 439. Letter. 16. Eachempati SR, Vaslef SN, Sebastian MW, Reed RL II. Blunt vascular injuries of the head and neck: is heparinization necessary? J Trauma. 1998;45:997–1004. 17. Willis BK, Greiner F, Orrison WW, Benzel EC. The incidence of vertebral artery injury after midcervical spine fracture or subluxation. Neurosurgery. 1994;34:435– 442. 18. Woodring JH, Lee C, Duncan V. Transverse process fractures of the cervical vertebrae: are they insignificant? J Trauma. 1993; 34:797– 802. 19. Zelenock GB, Kazmers A, Whitehouse WM Jr, et al. Extracranial internal carotid artery dissections: noniatrogenic traumatic lesions. Arch Surg. 1982;117:425– 432. 20. Biffl WL, Moore EE, Offner PJ, et al. Blunt carotid arterial injuries: implications of a new grading scale. J Trauma. 1999;47: 845– 853.
209 asymptomatic patients subsequently undergoing arteriography based on the new criteria presented today since 1996. In summary, 85 patients over a period of 7 or 8 years, an extraordinary number, were found to have injury to the cerebrovascular vessels at one hospital. As stated, they then used univariate logistic regression analysis based on incidences, followed by multiple logistic regression analysis using the most significant categorical variables and univariate analysis. The authors concluded that independent
THE AMERICAN JOURNAL OF SURGERY® VOLUME 178 DECEMBER 1999
521