Home > Browse Emergency Issues
<< Acute Spinal Injuries: Assessment and Management
ED Evaluation
Just as there is controversy as to whether someone needs immobilization after blunt trauma in the field, there is controversy about which of these immobilized patients require radiography to rule out a spine fracture. As mentioned previously, the cervical spine is the most delicate part of the spine and the most frequently injured. Due to this and the devastating consequence of ASCIs in the cervical spine, most radiological studies ordered by physicians evaluate this area. In 2000, an estimated 800,000 cervical spine radiographs were ordered in the US, and this number has likely increased since then.42 Because of the frequency of injuries requiring cervical spine evaluation, large studies that delineate decision rules to lower the number of these x-rays have been reported. Because the thoracic and lumbar spines are better supported and stronger than the cervical spine, the likelihood of injury is much lower. Thus, there is less need and fewer attempts to derive decision rules to decrease radiological studies of these areas.
Cervical spine
Clinicians tend to order imaging of the cervical spine for most patients with blunt head or neck trauma even if there is only a minute chance of injury. According to the National Hospital Ambulatory Medical Care Survey: Emergency Department Summary, the number of emergency department (ED) visits in the United States for mechanisms potentially causing spine injuries (falls, being struck, striking something, motor vehicle traffic incidents) was approximately 16.5 million in 2003.43 Only approximately 4% of all cervical spine radiographs reveal a fracture.44 Despite the low cost of x-rays, the sheer number of radiographs ordered contributes significantly to the financial burden of any health care system. This financial cost is in addition to the potential risks, significant discomfort, radiation exposure, and unnecessary space utilization in crowded emergency rooms caused by continued immobilization. To expedite the care of these patients, clinical decision rules have been developed to safely clear them from the cervical spine immobilization devices without the need for x-rays.
There are three important measures of such rules. To prevent missed cervical spine injuries and their consequences, it is imperative that such clinical decision rules have a high sensitivity as well as a high negative predictive value (NPV). The third important variable, although secondary to sensitivity and NPV, is the specificity of the rule. The higher the specificity of a decision rule, the greater number of unnecessary x-rays it will prevent. Investigators have derived two sets of criteria to evaluate the cervical spine in the last decade: the NEXUS Low-risk Criteria (NLC) and the Canadian
C-spine Rule (CCR).
The NEXUS was the first study to address the issue of eliminating x-rays for patients at low risk for injury. Prior to NEXUS, there were numerous small studies that suggested that patients with blunt trauma have a low risk of cervical spine injury if they met low-risk criteria.45‑54 Some of these studies reported a sensitivity of 100%. Because such sensitivities were calculated from studies with few patients, the lower confidence intervals for these sensitivities could be as low as 89%.48, 55 If this were the “true” sensitivity of the criteria, it would be too low to justify their general deployment to rule out cervical spine injury. In contrast, NEXUS was a much larger multicenter prospective observational study that tested five criteria to get a more accurate estimate of their sensitivity.
In this study, 34,069 patients who underwent radiography of the cervical spine following blunt trauma were evaluated. Of these patients, 818 had radiographically documented injuries. The study assessed the performance of a five-part decision rule to predict which patients would have a spinal injury (illustrated in Figure 11). All patients received either a 3-view cervical spine x-ray evaluation or a cervical spine computed tomography (CT) scan to determine if fractures were present. The sensitivity, specificity, and NPV of the NLC were calculated for all spine injuries that were considered to be significant. Insignificant injuries were defined as injuries that, if not identified, would be extremely unlikely to result in any harm to the patient. Of the significant injuries, the sensitivity and specificity of the test were 99.6% and 12.9%, respectively. The NPV was calculated to be 99.9%. Given this sensitivity, the miss rate for significant injuries translates into one missed injury in 17,000 patients. If the average emergency physician orders 50 cervical spine films a year, he/she would miss one significant injury every 340 practice years. Consequently, the NLC rapidly became the trusted rule of choice in emergency departments in the US.
In the United States there is a strong medico-legal incentive not to miss any spinal injuries and few financial disincentives to limit c-spine x-rays on patients with blunt trauma. This led to the widespread adoption of the NLC despite its relatively low specificity. However, the low specificity of the NLC posed problems for healthcare systems outside the US, leading to the development of the CCR.56 Outside of the US, the NLC were perceived as inefficient, and its application by different attendings varied widely.56 In Canada, for example, a much higher threshold for ordering c-spine x-rays already existed when the NLC were validated. After the NEXUS, there was concern there that adoption of the NLC might actually increase the number of x-rays ordered in the setting of blunt trauma.56 Therefore, Stiell et al set a goal to develop a rule with higher specificity that could decrease the number of x-rays ordered in Canada.
The CCR were derived from a set of clinical variables that were either strongly associated with injury or greatly decreased the likelihood of injury.56 The identified variables were combined into the CCR decision rule shown in Figure 12. This tool was prospectively evaluated against the NLC in 2003. The authors evaluated 8283 patients, 162 of whom had clinically important c-spine injuries. For “clinically important” injuries, the primary endpoint of the study, the sensitivity and specificity of the CCR were 99.4%, and 45.1%. In contrast with the sensitivity calculated in NEXUS, the NLC were found to have a sensitivity of only 90.7% on the patients in the CCR study. This translates into NPVs of 100% for the CCR and 99.4% for the NLC. Based on the CCR study results, the authors concluded that their criteria were not only more specific than the NEXUS criteria, but more sensitive as well (See Table 3). In a later 2004 study, Stiell et al go as far as to state that “the NEXUS low-risk criteria should be further explicitly and prospectively evaluated for accuracy and reliability before widespread clinical use outside the United States.” 57
Choosing a decision rule
The CCR study finding that the sensitivity of the NLC was only 90.7% contrasts dramatically with the sensitivity found in NEXUS of 99.6%. Certain factors may have been responsible for this.58 The Canadian team had an important methodological difference from NEXUS in that the Canadian group used a clinical follow-up protocol to evaluate blunt trauma patients in whom x-rays were deemed unnecessary. This sub-population totaled nearly 30% of all patients in the study. In contrast, the NEXUS investigators excluded these patients, and only examined patients who underwent x-rays. This difference is crucial, because it is very likely that the decision not to perform x-rays was guided (consciously or unconsciously) by the NLC.59 The NLC were already in widespread use during the CCR study, and the decision to deem patients safe for discharge without x-rays was probably impacted by this familiarity. By the inadvertent pre-selection of the study group, they eliminated some number of true negatives and false negatives upon which the rules were tested. The reduction in the number of false negatives will deceptively elevate the CCR’s sensitivity in comparison to the NLC. In addition, by including patients who may have not received x-rays in NEXUS, the CCR significantly increased the percentage of true negative patients. This will elevate both criteria’s specificity compared to what their specificities would be if they followed the NEXUS methodology. Indeed, this effect is exemplified in the CCR study by the unexpectedly high NLC specificity of 36.8%, compared to the 12.9% specificity seen in NEXUS. The lower sensitivity and higher specificity for the NLC found in the CCR study is consistent with the effects of these biases.59
Another potential flaw of the CCR study was the addition of clarifying definitions to the definitions of “intoxication” and “distracting injuries”. The NEXUS investigators deliberately left these definitions broad. In the case of distracting injuries, NEXUS investigators stated that “no precise definition for distracting injury is possible.” Consequently the investigators only give examples of such injuries. The CCR study’s use of surrogate NLC criteria will inherently cause misclassification errors and alter the performance of the NLC.44, 60 There were also differing patient eligibility criteria. Whereas NEXUS included patients of 16 years of age and under as well as people with a Glasgow Coma Score (GCS) of 15, the CCR study excluded them. A final important issue to recognize is that the CCR prospective study was performed in the same institutions from which the criteria were derived. Regional familiarity with the CCR rules probably existed and potentiates the possibility of bias. It also brings into question the CCR’s applicability to outside institutions.44
Traditional x-rays have long been the means for evaluating the spine after blunt trauma. For the cervical
spine, the NLC and CCR recommend that x-rays be performed on those that fail their low-risk criteria. Recommended radiographic examination of the cervical spine usually consists of a three-view x-ray: the lateral, anterior-posterior, and open mouth views, although five-view variations exist.61 A swimmer’s view is recommended if the top of T1 is not visualized. If the x-rays are deemed inadequate, a CT scan of the cervical spine is recommended.
In our opinion, the NLC and CCR clinical decision rules are roughly equivalent in their sensitivities. Both the NLC and the CCR were validated in prospective, randomized, multicenter trials, and we consider their validation studies Class I evidence. Our recommendation is that either one of the rules be used, and consider this the standard of care. We do not consider the lower sensitivity of the NLC determined in the CCR study to be an accurate representation of the NLC performance. X-rays should be obtained on patients who fail low-risk criteria.
Thoracolumbar spine
The prevalence of thoracic and lumbar spine injuries is 2-3% in blunt trauma victims. Although more rare, approximately 40-50% of these injuries are associated with a neurologic deficit, likely because of the tremendous forces needed to fracture this area of the spine.8 Figure 13 shows a MDCT reconstruction of such an injury, demonstrating a severe T4 on T5 spine fracture-dislocation caused by a high speed motor vehicle accident. In contrast to the numerous studies that have investigated clinical guidelines to evaluate the cervical spine, there are relatively few that guide the evaluation of the thoracic and lumbar spines.62-70 The studies that exist are retrospective reviews and essentially extrapolate the cervical spine data to these regions.71 Several studies demonstrate variables that are associated with spine fractures (listed in Table 4), and identify criteria similar to those used in the NLC or CCR criteria. A decision rule developed by Hsu et al was based on published factors shown to be retrospectively associated with thoracolumbar injury.65 A modification of his proposed clinical pathway is shown in Figure 14. This protocol was evaluated by a retrospective chart review in two groups of 100 patients: patients with confirmed thoracolumbar fractures, and randomly selected multi-trauma patients. Hsu reported a sensitivity of 100%, an NPV of 100%, and specificity of 11.3%. As with NEXUS, this specificity is quite low, and thus may cause implementation issues in regions that have a more restrictive threshold for obtaining radiological studies. The most sensitive of the criteria tested was found to be the combination of either back pain or midline tenderness, with a sensitivity of 62.1%. This association has been confirmed by other studies as well.70 The most specific criterion, not unexpectedly, was a palpable step-off over the spine, found to have a specificity of 100%.65 Another factor that should raise one’s index of suspicion for a fracture is a previously identified spine fracture (of the cervical spine, for example).72 This clinical decision rule awaits prospective validation.
Our opinion is that there is very little evidence to support a clinical decision rule for the evaluation of the thoracic and lumbar spines. The level of evidence is Class III. One should maintain a high index of suspicion in high-energy or multi-trauma patients. One spine fracture is often associated with another. A GCS<15 should also increase the level of diligence to rule out fractures of the thoracic and lumbar spine. Until a prospective validation study is performed on a clinical decision rule that assesses these areas of the spine, we recommend looking for the high-risk factors listed in Table 4 and using appropriate clinical judgment. If there is any likelihood of injury, radiographic studies are recommended.
Cervical spine
Clinicians tend to order imaging of the cervical spine for most patients with blunt head or neck trauma even if there is only a minute chance of injury. According to the National Hospital Ambulatory Medical Care Survey: Emergency Department Summary, the number of emergency department (ED) visits in the United States for mechanisms potentially causing spine injuries (falls, being struck, striking something, motor vehicle traffic incidents) was approximately 16.5 million in 2003.43 Only approximately 4% of all cervical spine radiographs reveal a fracture.44 Despite the low cost of x-rays, the sheer number of radiographs ordered contributes significantly to the financial burden of any health care system. This financial cost is in addition to the potential risks, significant discomfort, radiation exposure, and unnecessary space utilization in crowded emergency rooms caused by continued immobilization. To expedite the care of these patients, clinical decision rules have been developed to safely clear them from the cervical spine immobilization devices without the need for x-rays.
There are three important measures of such rules. To prevent missed cervical spine injuries and their consequences, it is imperative that such clinical decision rules have a high sensitivity as well as a high negative predictive value (NPV). The third important variable, although secondary to sensitivity and NPV, is the specificity of the rule. The higher the specificity of a decision rule, the greater number of unnecessary x-rays it will prevent. Investigators have derived two sets of criteria to evaluate the cervical spine in the last decade: the NEXUS Low-risk Criteria (NLC) and the Canadian
C-spine Rule (CCR).
The NEXUS was the first study to address the issue of eliminating x-rays for patients at low risk for injury. Prior to NEXUS, there were numerous small studies that suggested that patients with blunt trauma have a low risk of cervical spine injury if they met low-risk criteria.45‑54 Some of these studies reported a sensitivity of 100%. Because such sensitivities were calculated from studies with few patients, the lower confidence intervals for these sensitivities could be as low as 89%.48, 55 If this were the “true” sensitivity of the criteria, it would be too low to justify their general deployment to rule out cervical spine injury. In contrast, NEXUS was a much larger multicenter prospective observational study that tested five criteria to get a more accurate estimate of their sensitivity.
In this study, 34,069 patients who underwent radiography of the cervical spine following blunt trauma were evaluated. Of these patients, 818 had radiographically documented injuries. The study assessed the performance of a five-part decision rule to predict which patients would have a spinal injury (illustrated in Figure 11). All patients received either a 3-view cervical spine x-ray evaluation or a cervical spine computed tomography (CT) scan to determine if fractures were present. The sensitivity, specificity, and NPV of the NLC were calculated for all spine injuries that were considered to be significant. Insignificant injuries were defined as injuries that, if not identified, would be extremely unlikely to result in any harm to the patient. Of the significant injuries, the sensitivity and specificity of the test were 99.6% and 12.9%, respectively. The NPV was calculated to be 99.9%. Given this sensitivity, the miss rate for significant injuries translates into one missed injury in 17,000 patients. If the average emergency physician orders 50 cervical spine films a year, he/she would miss one significant injury every 340 practice years. Consequently, the NLC rapidly became the trusted rule of choice in emergency departments in the US.
In the United States there is a strong medico-legal incentive not to miss any spinal injuries and few financial disincentives to limit c-spine x-rays on patients with blunt trauma. This led to the widespread adoption of the NLC despite its relatively low specificity. However, the low specificity of the NLC posed problems for healthcare systems outside the US, leading to the development of the CCR.56 Outside of the US, the NLC were perceived as inefficient, and its application by different attendings varied widely.56 In Canada, for example, a much higher threshold for ordering c-spine x-rays already existed when the NLC were validated. After the NEXUS, there was concern there that adoption of the NLC might actually increase the number of x-rays ordered in the setting of blunt trauma.56 Therefore, Stiell et al set a goal to develop a rule with higher specificity that could decrease the number of x-rays ordered in Canada.
The CCR were derived from a set of clinical variables that were either strongly associated with injury or greatly decreased the likelihood of injury.56 The identified variables were combined into the CCR decision rule shown in Figure 12. This tool was prospectively evaluated against the NLC in 2003. The authors evaluated 8283 patients, 162 of whom had clinically important c-spine injuries. For “clinically important” injuries, the primary endpoint of the study, the sensitivity and specificity of the CCR were 99.4%, and 45.1%. In contrast with the sensitivity calculated in NEXUS, the NLC were found to have a sensitivity of only 90.7% on the patients in the CCR study. This translates into NPVs of 100% for the CCR and 99.4% for the NLC. Based on the CCR study results, the authors concluded that their criteria were not only more specific than the NEXUS criteria, but more sensitive as well (See Table 3). In a later 2004 study, Stiell et al go as far as to state that “the NEXUS low-risk criteria should be further explicitly and prospectively evaluated for accuracy and reliability before widespread clinical use outside the United States.” 57
Choosing a decision rule
The CCR study finding that the sensitivity of the NLC was only 90.7% contrasts dramatically with the sensitivity found in NEXUS of 99.6%. Certain factors may have been responsible for this.58 The Canadian team had an important methodological difference from NEXUS in that the Canadian group used a clinical follow-up protocol to evaluate blunt trauma patients in whom x-rays were deemed unnecessary. This sub-population totaled nearly 30% of all patients in the study. In contrast, the NEXUS investigators excluded these patients, and only examined patients who underwent x-rays. This difference is crucial, because it is very likely that the decision not to perform x-rays was guided (consciously or unconsciously) by the NLC.59 The NLC were already in widespread use during the CCR study, and the decision to deem patients safe for discharge without x-rays was probably impacted by this familiarity. By the inadvertent pre-selection of the study group, they eliminated some number of true negatives and false negatives upon which the rules were tested. The reduction in the number of false negatives will deceptively elevate the CCR’s sensitivity in comparison to the NLC. In addition, by including patients who may have not received x-rays in NEXUS, the CCR significantly increased the percentage of true negative patients. This will elevate both criteria’s specificity compared to what their specificities would be if they followed the NEXUS methodology. Indeed, this effect is exemplified in the CCR study by the unexpectedly high NLC specificity of 36.8%, compared to the 12.9% specificity seen in NEXUS. The lower sensitivity and higher specificity for the NLC found in the CCR study is consistent with the effects of these biases.59
Another potential flaw of the CCR study was the addition of clarifying definitions to the definitions of “intoxication” and “distracting injuries”. The NEXUS investigators deliberately left these definitions broad. In the case of distracting injuries, NEXUS investigators stated that “no precise definition for distracting injury is possible.” Consequently the investigators only give examples of such injuries. The CCR study’s use of surrogate NLC criteria will inherently cause misclassification errors and alter the performance of the NLC.44, 60 There were also differing patient eligibility criteria. Whereas NEXUS included patients of 16 years of age and under as well as people with a Glasgow Coma Score (GCS) of 15, the CCR study excluded them. A final important issue to recognize is that the CCR prospective study was performed in the same institutions from which the criteria were derived. Regional familiarity with the CCR rules probably existed and potentiates the possibility of bias. It also brings into question the CCR’s applicability to outside institutions.44
Traditional x-rays have long been the means for evaluating the spine after blunt trauma. For the cervical
spine, the NLC and CCR recommend that x-rays be performed on those that fail their low-risk criteria. Recommended radiographic examination of the cervical spine usually consists of a three-view x-ray: the lateral, anterior-posterior, and open mouth views, although five-view variations exist.61 A swimmer’s view is recommended if the top of T1 is not visualized. If the x-rays are deemed inadequate, a CT scan of the cervical spine is recommended.
In our opinion, the NLC and CCR clinical decision rules are roughly equivalent in their sensitivities. Both the NLC and the CCR were validated in prospective, randomized, multicenter trials, and we consider their validation studies Class I evidence. Our recommendation is that either one of the rules be used, and consider this the standard of care. We do not consider the lower sensitivity of the NLC determined in the CCR study to be an accurate representation of the NLC performance. X-rays should be obtained on patients who fail low-risk criteria.
Thoracolumbar spine
The prevalence of thoracic and lumbar spine injuries is 2-3% in blunt trauma victims. Although more rare, approximately 40-50% of these injuries are associated with a neurologic deficit, likely because of the tremendous forces needed to fracture this area of the spine.8 Figure 13 shows a MDCT reconstruction of such an injury, demonstrating a severe T4 on T5 spine fracture-dislocation caused by a high speed motor vehicle accident. In contrast to the numerous studies that have investigated clinical guidelines to evaluate the cervical spine, there are relatively few that guide the evaluation of the thoracic and lumbar spines.62-70 The studies that exist are retrospective reviews and essentially extrapolate the cervical spine data to these regions.71 Several studies demonstrate variables that are associated with spine fractures (listed in Table 4), and identify criteria similar to those used in the NLC or CCR criteria. A decision rule developed by Hsu et al was based on published factors shown to be retrospectively associated with thoracolumbar injury.65 A modification of his proposed clinical pathway is shown in Figure 14. This protocol was evaluated by a retrospective chart review in two groups of 100 patients: patients with confirmed thoracolumbar fractures, and randomly selected multi-trauma patients. Hsu reported a sensitivity of 100%, an NPV of 100%, and specificity of 11.3%. As with NEXUS, this specificity is quite low, and thus may cause implementation issues in regions that have a more restrictive threshold for obtaining radiological studies. The most sensitive of the criteria tested was found to be the combination of either back pain or midline tenderness, with a sensitivity of 62.1%. This association has been confirmed by other studies as well.70 The most specific criterion, not unexpectedly, was a palpable step-off over the spine, found to have a specificity of 100%.65 Another factor that should raise one’s index of suspicion for a fracture is a previously identified spine fracture (of the cervical spine, for example).72 This clinical decision rule awaits prospective validation.
Our opinion is that there is very little evidence to support a clinical decision rule for the evaluation of the thoracic and lumbar spines. The level of evidence is Class III. One should maintain a high index of suspicion in high-energy or multi-trauma patients. One spine fracture is often associated with another. A GCS<15 should also increase the level of diligence to rule out fractures of the thoracic and lumbar spine. Until a prospective validation study is performed on a clinical decision rule that assesses these areas of the spine, we recommend looking for the high-risk factors listed in Table 4 and using appropriate clinical judgment. If there is any likelihood of injury, radiographic studies are recommended.
Sign up for free access