The majority of cervical spinal cord injury occurs from unstable cervical spine fracture. Unfortunately, unstable cervical spine fractures may be clinically unapparent upon presentation to the emergency department. Only one third of spinal trauma patients present initially with a neurological deficit (6, 7). Moreover, important clinical features such as pain from injury may be masked by other injuries, medication, and drug and alcohol intoxication. Patients who have spine fractures but are neurologically intact at the time of presentation to the emergency department may progress and develop neurologic compromise (8, 9 and 10). In an early study by Rogers (11) in the 1950s, progression to neurologic deficit occurred in 50% of subjects with cervical spine fracture who were initially neurologically intact. With modern field and emergency department spine immobilization techniques, this rate is much lower. However, based on a longitudinal study of 253 spine fracture patients in Alberta, Reid reported that progression to neurologic deficit in patients with initially
missed cervical spine fractures is as high as 10% (8). Accordingly, the current consensus recommendations of the American College of Surgeons Advanced Trauma Life Support (12) and the Appropriateness Criteria of the American College of Radiology (13) advise imaging of the spine in patients who are at risk for fracture.

Defining the group of subjects who are at risk for cervical spine fracture and therefore in whom imaging is appropriate remains challenging. Cervical spine imaging is one of the most common imaging exams performed on trauma patients in many trauma centers throughout the United States and other developed nations. The yield from such imaging however is very low with only 0.9% to 2.8% of imaging studies demonstrating fractures (14, 15 and 16). Although cervical spine imaging itself is not an expensive procedure, the frequency of performance makes it a high-cost item for trauma care (17, 18).

Efforts to define optimal imaging selection criteria have centered on the development and validation of clinical prediction rules. Clinical prediction rules are decision-making tools consisting of several factors that suggest a course of action or provide a probability of a disease or injury (19). Clinical prediction rules composed of clinical factors can be used to determine the probability of injury or can be applicable in selecting subjects appropriately for imaging. In order to be useful, clinical prediction rules must be easy to use, strongly predictive, and yield consistent results with multiple users and diverse populations. The process of determining a clinical prediction rule involves developing a possible set of clinical criteria, testing interobserver reliability, and validating the clinical prediction rule in a second set of subjects (19, 20).

In the past decade, two large prospective multicenter investigations have been performed to define clinical prediction rules for selection of appropriate subjects for cervical spine imaging after trauma. The first of these was the National Emergency X-Ray Utilization Study (NEXUS) in the United States, published in 2000 (15). The NEXUS study was a prospective validation of clinical criteria that were already in place at emergency departments throughout the United States. The NEXUS investigators developed a collaboration of 21 emergency departments and enrolled a total of 34,069 trauma subjects in whom imaging was requested. Eight hundred eighteen (2.4%) of the subjects had cervical spine fractures. The NEXUS study group applied an a priori prediction rule of clinical criteria that specified when imaging was indicated, and it used the radiologist’s interpretation of the imaging results as the gold standard for evaluating the effectiveness of the criteria. The NEXUS criteria for not imaging (Table 4-1) included normal neurologic examination, absence of posterior midline tenderness, and a normal level of alertness without intoxication or painful distracting injuries. The NEXUS study did not affect the ongoing clinical care of the study subjects, so imaging consisted primarily of radiography, supplemented by computed tomography (CT) and magnetic resonance imaging (MRI) as dictated by the clinical status of each patient and by local practice patterns. The NEXUS group found that their clinical prediction rule could adequately identify subjects at risk of fracture, with a sensitivity of 99.6%. In addition, interobserver agreement was excellent (kappa 5 = 0.73) (15).

Unfortunately, while the sensitivity of the NEXUS criteria was high, the specificity was rather low (12.6%), indicating that the number of unnecessary normal radiographic studies that use of the NEXUS rule would eliminate was small (15). Because the NEXUS study relied principally on imaging criteria that pre-existed at participating centers, there was no measurable impact on imaging utilization.

Important strengths of the NEXUS study were its large sample size and inclusion of multiple emergency departments from a range of sizes and types of facilities, including academic centers, community hospitals, level-one trauma centers, as well as low-volume trauma centers. The broad representation in the study sample suggests that, at least in the United States, the guidelines set forth by the NEXUS study can be applied to a wide range of hospital settings.

Subsequent to the NEXUS study, the Canadian cervical spine study group performed their own separate validation of clinical prediction rules in determining when imaging is appropriate in cervical spine trauma (21). The first phase of the Canadian study was the prospective development of a clinical prediction rule from 8,924 subjects who were evaluated at ten hospitals in Ontario. Investigators used 20 unique clinical predictors of cervical spine injury that were not commonly relied on in standard clinical practice. The Canadian study differs from the NEXUS study primarily due to its usage of different clinical parameters to qualify appropriate usage of imaging in cervical spine trauma. From this initial development study, the Canadian investigators identified criteria that had 100% sensitivity and 42.5% specificity for detection of acute cervical spine injury (Table 4-2). The second
portion of the Canadian study was a validation of the criteria developed in the first portion of the study. This was performed in the same ten hospitals on another 8,283 trauma patients. The results of this validation study found that the clinical prediction rule had a 99.4% sensitivity and a 45.1% specificity for predicting when a cervical spine injury was present with the additional benefit of a high interobserver agreement (14, 21).

It is difficult to draw direct comparisons between the NEXUS and the Canadian Cervical Spine results because of differences in the study populations on which they were validated. The Canadian study only included subjects who had a Glasgow coma scale (GCS) score of 15 (i.e., normal) and who were hemodynamically stable (21). The NEXUS included trauma patients regardless of GCS scores and hemodynamic instability, although the decreased mental status was one of the criteria for performance of radiography (15, 22). In addition, the Canadian study included all subjects in whom cervical spine injury was suspected due to mechanism and other defined clinical criteria with the treating physician’s decision to order radiography having no bearing on enrollment (21), while NEXUS was limited to subjects who underwent imaging with the decision to perform radiography left to the physician’s discretion (15, 22).

Perhaps the most important difference between the two studies is the specificity of the prediction rules. The higher specificity of the Canadian C-Spine Rule (42.5%), as compared to the NEXUS rule (12.9%), implies that the Canadian C-Spine Rule may have a greater impact on utilization by decreasing unnecessary ordering of cervical radiographs by a greater proportion. However, this higher specificity may also be explained by differences in study group selection between the Canadian and NEXUS cohort (23). To date, no data have been published on the actual effect of either rule on imaging ordering behavior in practice or on any changes in the performance of unnecessary imaging. A concern with the Canadian cervical spine rule is that the criteria are perceived to be much more complex than the NEXUS rule, and it includes provocative testing for symptoms (e.g., head turning and flexion), a practice with which some US physicians may not be comfortable. A lively debate has emerged between the two research groups, but no clear advantage of one clinical prediction rule over the other has been demonstrated (14, 23, 24, 25 and 26). In an attempt to provide standards for practice, the American Association of Neurological Surgeons/Congress of Neurological Surgeons (AANS/CNS) reviewed the literature and recommended that radiographic assessment is not necessary for those asymptomatic patients who are “awake, alert, and not intoxicated, who are without neck pain or tenderness, and who do not have significant associated injuries that detract from their general evaluation” (27).

Subjects identified as at risk for cervical spine imaging by either the NEXUS or Canadian cervical spine rule represent a large and heterogeneous group. Further, different imaging technologies may be appropriate in different patients. The traditional initial approach for evaluation of the acutely injured cervical spine has been with radiography. As of early 2005, the three-view radiography series (anteroposterior, lateral, and open mouth) is still the imaging modality of choice recommended by the American College of Radiology Appropriateness Criteria (13) and the Advanced Trauma Life Support (ATLS) course of the American College of Surgeons (12). The AANS/CNS recommends this three-view radiograph series as the initial study for symptomatic patients (27). Based on a critical literature review in 1997, the pooled estimate of the sensitivity of traditional cervical radiography for fracture is 94% (28). However, this sensitivity may be much lower in major multisystem trauma victims (29, 30). In addition, the specificity is affected by the characteristics of the patient being imaged (31). In patients with major trauma and multiple injuries, the probability of obtaining adequate cervical spine radiographs of diagnostic quality is diminished. For example, one study found that in subjects
with head injury, adequate cervical spine radiographs will only be obtained in 89% of patients, and in subjects with head injury from high-energy mechanisms such as motor vehicle and motorcycle crashes, adequate radiographs can only be obtained 78% to 84% of the time (31). In addition, in a study of multitrauma, intensive care unit patients in Alberta, Canada, radiographs were inadequate in 82% (32). Factors that degrade the quality of radiographs include the use of devices that obscure portions of the spine and soft tissues such as backboards or endotracheal tubes, the presence of other immobilizing injuries (such as upper extremity fractures precluding use of the swimmer’s view), and impairment of the patient’s cognitive status from drugs, cerebral injury, or hypoxia. The continued value of cervical radiography in the primary evaluation of trauma patients is due its wide availability at virtually all trauma centers, as well as the widespread experience in interpretation of radiography.

Use of helical CT scanning, first proposed by Núñez et al. (33, 34) in the mid-1990s, has gained acceptance as a more accurate alternative to radiography for initial evaluation of the cervical spine (29, 32, 35, 36). Single detector CT scan has a sensitivity of 98% to 99% for fracture with a specificity of 93% (29, 34, 37). In addition to being more accurate, CT of the cervical spine may also be acquired more rapidly than radiography when performed on subjects who are to undergo CT of other body regions. In patients with minor injury, radiography may require only 10 minutes on average, but this time increases up to an hour in subjects with multiple injuries from major trauma (17). Núñez et al. (33) reported that time in the emergency department for major trauma patients was decreased when CT was used as the initial cervical spine imaging approach. Subsequent investigations have supported the time advantage of CT (35, 38).

Blackmore et al. (19, 28, 37, 39) provide a definition of appropriate subjects for use of CT through a combination of a clinical prediction rule and cost effectiveness analysis. Cost-effectiveness analysis is a health policy tool that balances the outcomes and dollar costs of differing approaches to patient care. Cost-effectiveness analysis is usually based on a theoretical computer model that includes all of the possible outcomes from a medical care decision, with the associated short- and long-term costs (40, 41). For the cervical spine, the cost-effectiveness model incorporates the choice between CT and radiography, with the probability that an injury is present, the probability that the injury will be missed by each imaging approach, and the effect on the patient of missed injury. In addition, the model considers the costs of the initial imaging, the costs of any further imaging tests induced by false positive studies, and the costs of any adverse outcomes resulting from missed injury (28).