Quality care of trauma patients depends on an efficient and systematic approach for correctly diagnosing clinically important injuries. Spinal injuries can be devastating, with 19% to 50% of thoracolumbar spine fractures reportedly being associated with neurological injury1-3. For asymptomatic patients with negative findings on clinical examination, normal mental status, and no distracting injuries, no radiographic screening is required4-6. However, for patients who present to the emergency department with any form of altered mental status, including those who are inebriated or intubated, the clinician must rely on radiographic imaging to detect major injuries4,6-11. It is accepted practice that cervical spine injuries can be effectively detected with use of computerized tomographic scans of the cervical spine, eliminating the use of conventional radiographs12,13. Thus, the question arises: does a nonreconstructed computerized tomographic scan of the abdomen and pelvis or chest that is used to evaluate intra-abdominal or thoracic injuries have enough musculoskeletal detail to rule out thoracic or lumbar spine fractures without additional tests?
In terms of assessing injuries to the thoracolumbar spine, the current gold standard diagnostic method is the use of anteroposterior and lateral radiographs dedicated to the thoracolumbar spine14,15. The existing Advanced Trauma Life Support guidelines require blunt-force trauma patients who present with altered mental status to undergo computerized tomography of the abdomen and pelvis in order to rule out visceral injuries; however, this imaging study is not meant for the detection of thoracolumbar spine fractures16. However, technology has improved so that radiologists can now view these scans with specific filtering to optimally view bone as well.
The evolution of multidetector computed tomography technology has revolutionized imaging capabilities. Many recent studies have suggested that reformatted multidetector computed tomography provides adequate information to detect clinically important thoracolumbar spine injuries1,17-20. Studies have shown that decreased time, cost, and radiation dose are incurred with the implementation of multidetector computerized tomography protocols, eliminating the role of radiographs in screening for thoracolumbar spine injuries1,20. The most recent update on screening for thoracolumbar spine fractures suggested that multidetector computerized tomographic reformatted images are superior21. However, this recommendation is vague and does not clearly define which radiographic views and studies should be utilized for evaluating the spine in trauma patients. Clear definitions of the radiographic studies need to be provided to further establish protocols for evaluating the thoracolumbar spine.
Three imaging modalities currently exist to assess the spine: (1) anteroposterior and lateral radiographs, (2) computerized tomography of the chest, abdomen, and pelvis for the evaluation of visceral pathology, and (3) computerized tomographic reconstructions dedicated to the spine. In an effort to identify reliable radiographic modalities for the detection and evaluation of thoracolumbar fractures, we propose two hypotheses. First, nonreconstructed computerized tomography has a higher sensitivity and specificity for the detection of thoracolumbar spine injuries than plain radiography does. Second, nonreconstructed computerized tomography of the abdomen and pelvis with 5-mm slices is a reliable screening tool for thoracolumbar spine injuries in blunt trauma patients with altered mental status.
Patient Population
Fifty-nine consecutive patients who presented with altered mental status after blunt trauma were admitted to the trauma service at Parkview Hospital (an American College of Surgeons Level-II trauma center) and were enrolled into this prospective cohort study. Institutional review board approval was obtained with approval of our informed consent procedure. All patients who were evaluated with a nonreconstructed computerized tomographic scan of the abdomen and pelvis and anteroposterior and lateral radiographs of the thoracic and lumbar spine were eligible for inclusion in the study. Altered mental status was defined as inebriation, intubation at the scene or in the emergency department prior to clinical examination by a trauma surgeon, confusion or repetitive speech, and/or obtundation or seizure activity. This definition included all patients in whom the physical examination findings were unreliable and for whom imaging was required in order to screen reliably for spine injuries. As per Advanced Trauma Life Support guidelines, patients with altered mental status as described above were evaluated with anteroposterior and lateral thoracolumbar radiographs as a screening tool for spine fracture16. Patients underwent computerized tomography of the head, cervical spine, and abdomen and pelvis as determined by the trauma surgeon and according to Advanced Trauma Life Support guidelines. (Computerized tomography of the chest was ordered in the presence of clinical or radiographic evidence of substantial thoracic or great vessel injury; therefore, it was not used for all patients.) Data such as the mechanism of injury, the Injury Severity Score, sex, and age were collected from the trauma registry. The Injury Severity Score is an anatomical scoring system designed to provide an overall score for trauma patients with multiple injuries. The Injury Severity Score is the sum of squares of the three highest abbreviated injury scale scores for injuries of different body regions. Injury Severity Scores range from 1 to 75.
Radiographic Protocol and Imaging Definitions
Plain anteroposterior and lateral radiographs of the thoracolumbar spine were made with a portable radiography machine in the traditional fashion. The nonreconstructed computerized tomographic scan of the abdomen and pelvis and of the chest (when clinically indicated) was completed with a GE LightSpeed 16 multidetector computerized tomography scanner (GE Healthcare, Waukesha, Wisconsin) with volume acquisition post-processed at 5-mm axial collimation, a pitch of 1.375 to 1.000, and thick-slab coronal reformats. In cases in which several trauma patients were in need of computerized tomographic scans simultaneously, the GE VCT-64 slice scanner (GE Healthcare) was used at the same settings for all patients. This was done at the discretion of the attending trauma surgeon when it was believed that patient care would be compromised by having each patient wait to undergo computerized tomography with use of the scanner in the emergency department instead of being transported to another available scanner in the radiology department. Reconstructions of the spine were defined as =2-mm (generally 0.625 or 1.25-mm) collimation thin-slice scans dedicated (i.e., zoomed and targeted) to the spine in the axial, sagittal, and coronal planes. Computerized tomographic reconstructions dedicated to the spine were completed for all patients and were used as the so-called gold standard for calculating sensitivity and specificity. These reconstructions were derived by post-processing the raw volume data set, without the need to rescan the patient. (Thus, the patient's clinical evaluation and treatment were not delayed or disrupted and the patient incurred no additional radiation dose.) For the purpose of the present study, a so-called Pan-scan was defined as a single, nonreconstructed computerized tomographic scan of the chest, abdomen, and pelvis.
Interpreting Radiographic Studies
All studies that had been ordered were evaluated by the staff radiologists at the time of the injury, and formal written interpretations were rendered immediately with use of voice-recognition dictation software. Results were systematically logged by the author (M.W.S.). A second radiologist (J.D.R.), who was not the original interpreter of any of the studies and was blinded to the results, reviewed and reinterpreted all radiographic examinations. Each imaging modality was reviewed separately to minimize bias in interpretation. In cases of discrepancies between the original interpretation and the interpretation by the second radiologist, a neuroradiologist (R.F.), who also did not provide the original report, made the final diagnosis.
Determination of Fracture
Fractures that were identified on imaging studies were evaluated for cortical disruption of the bone and for evidence of acuity, such as the presence of hematoma in the area or a lack of sclerosis. The reconstructions were implemented as the most sensitive imaging modality for the detection of a fracture. A follow-up clinical examination was completed for all patients by one of us (M.W.S.) once the patient's mental status had cleared to ensure that the radiographic evidence matched the clinical findings. This follow-up clinical examination was accomplished for all patients except one, who died in the hospital secondary to severe head trauma and multiple-system organ failure. This patient had had previous spine fractures, and so the new images were compared with previous studies to ensure that the acute fractures attributed to the current injury were indeed new and were not present on the previous computerized tomographic scans. In all cases, patients who were declared to have a fracture had positive radiographic and physical examination findings and a discharge diagnosis of a spine fracture. Positive physical examination findings included point tenderness to palpation of the posterior elements of the spine or focal unilateral tenderness to deep palpation of the paraspinous muscles, indicative of a transverse process fracture.
Statistical Analysis
Sensitivity and specificity were calculated with use of the reconstructed computerized tomographic scan dedicated to the spine as the gold standard for both radiographs and nonreconstructed computerized tomography. The data analysis was divided into specific analyses with several subgroups: (1) thoracic spine fractures, (2) lumbar spine fractures, (3) fractures within the anatomical range for computerized tomography of the abdomen and pelvis and computerized tomography of the chest, and (4) thoracic spine fractures, lumbar spine fractures, and all spine fractures in patients receiving the Pan-scan.
For the anatomical range analysis, computerized tomography of the abdomen and pelvis was defined superiorly by the vertebral level associated with the apex of the diaphragm and inferiorly by L5. Computerized tomography of the chest was defined superiorly by T1 through the vertebral level associated with the inferior border of the thoracic cavity. Data analysis was conducted to calculate the sensitivity, specificity, positive predictive value, and negative predictive value for each individual fracture detected. Each vertebra was designated as having only one type of fracture, with the more clinically important fracture in terms of stability being logged. For example, if a vertebra had a burst fracture and also a fracture involving the transverse processes, it was logged as having a burst fracture only. In addition, analysis was performed at a patient level, with the patient being evaluated for the presence or absence of a fracture. In the former subset analysis, each fracture was counted. In the latter situation, multiple fractures in a single patient were counted as one fracture and were counted even in the event that the imaging modality missed a fracture but still identified others that were present in that patient. For all analyses, 2 × 2 tables were completed with the reconstructed computerized tomographic scan dedicated to the spine being used as the gold standard and with the test modality being either the nonreconstructed computerized tomographic scan or the radiograph. Sensitivity was defined as the number of true-positive results divided by the sum of true-positive and false-negative results. Specificity was defined as the number of true-negative results divided by the sum of true-negative and false-positive results. The negative predictive value was defined as the number of true-negative results divided by the sum of the true-negative and false-negative results. The positive predictive value was defined as the number of true-positive results divided by the sum of true-positive and false-positive results. DAG Stat (Victoria, Australia) was used to calculate 95% confidence intervals22. The level of significance was set at p < 0.05.
Source of Funding
No funding was received to support this study.
Patient Characteristics
Sixty-three patients met the screening criteria, but four were excluded from the study and from the data analysis: one patient died before radiographs were made, one was discharged before radiographs were made, one did not undergo computerized tomography of the abdomen and pelvis because of immediate transportation to the operating room for emergency surgery, and one presented with a computerized tomographic scan of the abdomen and pelvis that had been completed at a transferring institution. The average age (and standard deviation) of the remaining fifty-nine patients was 35 ± 15 years. The study group included forty-six male patients (78%) and thirteen female patients (22%).
Altered mental status often was due to one or more factors. A total of twenty-three patients were intubated prior to evaluation by a trauma surgeon. A total of seventeen patients were inebriated at the time of presentation (mean blood alcohol content, 0.24 gm%), with nine patients having inebriation as the only factor associated with the altered mental status. Twenty-six patients were confused or were speaking repetitively, with seven of them also being inebriated. One patient was obtunded or went into seizure activity on examination; this patient was also inebriated.
With regard to the mechanism of injury, forty-seven patients had been involved in a motor-vehicle collision (with thirty-one patients having been involved in an automobile or truck accident and sixteen having been involved in a motorcycle or moped accident), six patients had fallen an average of twenty feet (6.1 m) (range, ten to thirty feet [3.0 to 9.1 m]), four patients had been pedestrians who had been struck by a vehicle, and two patients had been victims of blunt assault.
For this patient population (n = 59), the mean Injury Severity Score (and standard deviation) was 17 ± 13. The mean Injury Severity Score was 23 ± 15 for patients with fracture (n = 21), compared with 13 ± 11 for those without fracture (n = 38) (p < 0.01). The mean Injury Severity Score was 27 ± 16 for patients with a lumbar spine fracture (n = 15), compared with 19 ± 10 for those with a thoracic spine fracture (n = 10); this difference was not significant. The mean Injury Severity Score for patients with both thoracic and lumbar spine fractures (n = 4) was 24 ± 8.
With regard to the type of screening performed, fifty-nine patients had anteroposterior and lateral thoracolumbar spine radiographs and computerized tomography of the abdomen and pelvis, thirty-five patients had computerized tomography of the chest, forty-one had reconstructions of the thoracic spine, and fifty-seven had reconstructions of the lumbar spine. Thirty-five patients had the Pan-scan, radiographs, and reconstructions of both the thoracic and lumbar spine. In twenty-one instances, there was a discrepancy for a particular imaging modality between the original reading and the reading by the independent radiologist. In twelve instances, it was simply a discrepancy in the number of transverse process fractures noted. In nine instances, there was a discrepancy in either the level or the number of end plate compression fractures noted. In all instances, the blinded neuroradiologist made the final determination.
Fracture Characteristics
Overall, the computerized tomographic reconstructions dedicated to the spine demonstrated seventy-two thoracic or lumbar fractures in twenty-four patients. Nonreconstructed computerized tomography of the abdomen and pelvis demonstrated fifty-eight fractures, and that of the chest demonstrated sixteen fractures. Forty-nine lumbar spine fractures were detected in fifteen patients, and twenty-three thoracic spine fractures were detected in ten patients. In the present study, twenty-four (41%) of the fifty-nine patients had a fracture that was demonstrated on the reconstructed computerized tomographic scan dedicated to the spine. There were forty-five transverse process fractures (63% of all fractures detected), accounting for eleven (48%) of twenty-three thoracic spine fractures and thirty-four (69%) of forty-nine lumbar spine fractures. There were five end plate fractures (including two in the thoracic spine and three in the lumbar spine), three laminar fractures (including one in the thoracic spine and two in the lumbar spine), seven spinous process fractures (including five in the thoracic spine and two in the lumbar spine), and five vertebral body compression fractures (including two in the thoracic spine and three in the lumbar spine). There were six burst fractures (including one in the thoracic spine and five in the lumbar spine); three of the lumbar spine burst fractures were at L4. There was one Chance fracture at T12.
Screening Sensitivity for Fractures
Computerized tomography of the abdomen and pelvis was more sensitive than plain radiographs for the detection of all fractures in both the thoracic spine (73% compared with 13%) and the lumbar spine (94% compared with 47%) (Table I). With regard to the detection of all vertebral level fractures in a patient, the nonreconstructed computerized tomographic scan was more sensitive for lumbar spine fractures than for thoracic spine fractures (94% compared with 73%). With regard to the detection of the presence of a lumbar spine fracture or thoracic spine fracture in a patient, the nonreconstructed computerized tomographic scan had sensitivities of 100% and 80%, respectively, whereas radiographs had sensitivities of 60% and 30%, respectively (Table I). These data were patient-level data (rather than fracture-level data), and therefore multiple fractures in a patient were counted as a single fracture. With regard to the evaluation of fractures specific to the range detected with computerized tomography of the abdomen and pelvis as well as of the chest, the results were similar (Table II). The nonreconstructed computerized tomographic scan had concordantly high specificity, negative predictive value, and positive predictive value and was superior to plain radiographs (Tables I and II).
When the data were isolated for patients who had all three studies (radiographs, Pan-scan, and thoracic spine and lumbar spine reconstructions), the findings were similar, with the Pan-scan being more sensitive for the detection of lumbar fractures than thoracic fractures (94% compared with 80%) when used to evaluate for all fractures. When used to evaluate the presence of a fracture in a patient, the sensitivity of the Pan-scan was 100% for lumbar spine fractures and 88% for thoracic spine fractures. When the analysis included all thoracic spine and lumbar spine fractures, the sensitivity of the Pan-scan was 89% for the detection of all fractures and 100% for the determination of the presence of a fracture (Table III). All patients who had thoracic spine fractures that were missed had lumbar spine fractures that were detected.
For the analysis of fractures that required intervention (bracing or surgery), such as compression, burst, or Chance fractures, the nonreconstructed computerized tomographic scan was superior to plain radiographs. The nonreconstructed computerized tomographic scan was 100% sensitive for thoracic and lumbar compression, burst, or Chance fractures. However, radiographs were only 75% sensitive for such fractures in the thoracic spine and were 63% sensitive for such fractures in the lumbar spine.
Failed Detection/Misclassification of Fractures
In the lumbar spine, three fractures were missed on the nonreconstructed computerized tomographic scan, including two transverse process fractures and one superior end plate fracture. These three fractures were in patients with other fractures that were detected on the nonreconstructed computerized tomographic scan. In the thoracic spine, six transverse process fractures in three patients were missed on the nonreconstructed computerized tomographic scan. One patient had three transverse process fractures that were missed but had other thoracic spine fractures that were detected on the nonreconstructed computerized tomographic scan. The second patient had two transverse process fractures that were missed and had no other fractures that were detected with the reconstructions; this patient was not part of the Pan-scan group. The third patient had a T2 transverse process fracture that was missed but had lumbar spine transverse process fractures that were detected with the nonreconstructed computerized tomographic scan as part of a Pan-scan. No fractures that were undetected with the nonreconstructed computerized tomographic scan required surgical or nonsurgical intervention. However, a T12 Chance fracture that required bracing was missed on standard radiographs and two burst fractures that required surgical instrumentation were misclassified as compression fractures on plain radiographs.
Our study group consisted of severely injured patients. The average Injury Severity Score of 17 was consistent with polytrauma and was slightly higher than that reported by Hauser et al.23 in a previous prospective analysis of trauma patients undergoing screening for thoracolumbar spine injuries, for whom the average Injury Severity Score was 12. The average Injury Severity Score for our cohort of patients with fractures was similar to that in the report by Sheridan et al.1, which focused specifically on patients with fractures (23 compared with 21). This finding suggests that, overall, our study population was equivalent to those in other studies, as the present study focused on patients who were at an increased risk for a spine fracture not being diagnosed secondary to their altered mental status.
We calculated the sensitivity and specificity for the detection of all fractures that were present as well as for the simple presence of a fracture. Nonreconstructed computerized tomography outperformed plain radiographs as a screening tool, with only minor fractures that did not require any additional intervention being missed. Our analysis also showed high specificity for the detection of fractures with nonreconstructed computerized tomography as well as concordant predictive values. These results are similar to other published data regarding the sensitivity of computerized tomography for the detection of spine fractures in trauma patients1,19,23,24. Whereas other studies have defined a fracture simply as a patient with a fracture1,23, the present study also analyzed all fractures that were present in a patient; no significant decrease in sensitivity was observed when evaluating for all fractures. The present study therefore provides a more thorough analysis of the reliability of nonreconstructed computerized tomography for the screening of thoracolumbar spine injuries as compared with other studies.
In the present study, we followed the current Advanced Trauma Life Support guidelines, ensuring that all patients received radiographs of the spine and computerized tomography of the abdomen and pelvis as well as computerized tomography of the chest when clinically indicated. On the basis of our data, there is sufficient evidence that radiographs of the lumbar spine can be omitted from the radiographic workup as long as computerized tomography of the abdomen and pelvis is completed with 5-mm slices, with reformats completed in the coronal plane. When computerized tomography of the chest is also completed, radiographs of the thoracic spine are not necessary to safely rule out clinically important thoracolumbar spine injuries. We therefore support a change in the imaging algorithm for blunt trauma patients with altered mental status.
At our institution, hemodynamically stable patients undergo computerized tomography of the abdomen and pelvis as well as of the chest when indicated, and they are evaluated with radiographs only when they are not stable enough to be safely transported to the scanner or when the anatomical region is not scanned. The entire process of post-processing the data adds time as the scanner technician must otherwise formulate the reconstructions and the staff radiologist and trauma surgeon must review the additional scan. Empirically, not relying on reconstructions to clear the spine radiographically saves approximately thirty to sixty minutes overall, depending on the patient load in the emergency department.
Whether reformats in the coronal plane, sagittal plane, or both need to be done is debatable but, regardless, one additional plane should be chosen to maintain the traditional radiologic principle of orthogonal views to assess alignment. Gestring et al.14 suggested using the axial computerized tomographic scan supplemented with the scout radiograph made for patient positioning to assess alignment in the sagittal plane. While reformatted computerized tomographic scans in the coronal plane are more beneficial for the general trauma surgeon when assessing visceral pathology, the orthopaedic surgeon may prefer to assess the spine in the sagittal plane to determine a discrepancy in interspinous process distance when assessing flexion-distraction injuries, also known as Chance fractures. In the present study, one osseous Chance fracture (1.4% of the detected fractures) was detected on nonreconstructed computerized tomography as well as the reconstructions and was missed on plain radiographs. It could be debated that a completely ligamentous Chance injury could be missed with the study protocol. However, depending on the study, flexion-distraction fractures make up only 3% to 15% of all spine fractures25,26. In addition, a purely ligamentous injury would be even more rare as one series of only Chance fractures did not include any ligamentous-only injuries among fifty-three cases26. Of note, Bernstein et al.26 found several findings to diagnose a Chance fracture on the axial or coronal view: widening of the facet joints; increased intercostal space; widening of the interpediculate distance (also suggestive of a burst component); horizontal fractures through the pedicles; transverse process, laminae, and articular process fractures with transverse fracture characteristics; and progressive loss of the pedicle on the axial view. With the rarity of this injury pattern, we recommend additional studies be conducted to adequately address this question.
A recent controversial topic is whether all patients presenting with altered mental status should simply have a Pan-scan for the radiographic evaluation, employing computerized tomography for imaging of the head, cervical spine, chest, abdomen, and pelvis. With newer imaging capabilities, significant increases in the utilization of computerized tomography of the chest have been observed27. In addition, the use of computerized tomography dedicated to the thoracolumbar spine has increased without clear reason18. The issues of efficiency, alleviating the need to order radiographs and to put the patient through another test, are obvious. Time is also saved as the technician will not have to stop and reformat scanning parameters for the patient. However, as physicians, we must also be cognizant of the added monetary cost and the exposure of our patients to radiation. Additional evidence is required to support this practice.
Despite the fact that the present study protocol complied with the STARD (Standards for the Reporting of Diagnostic accuracy studies) checklist28, it had several limitations. First, not all patients underwent a computerized tomographic scan of the chest. Even though all patients had radiographs of the thoracic spine, obviously there was a chance of missing a fracture as we found that plain radiographs had a low sensitivity for the detection of a fracture in the thoracic spine. However, we strictly followed current Advanced Trauma Life Support guidelines, and this further supports the consideration for routine computerized tomographic scanning of the chest, abdomen, and pelvis in blunt trauma patients presenting with an altered mental status. Even though we cannot fully account for any missed fractures in the thoracic spine not visualized with computerized tomography of the abdomen and pelvis or cervical spine, no patient was discharged with back pain or physical examination findings consistent with spine injury once the alteration in the mental status had cleared and no patient presented with complaints of back pain with findings suggestive of spine fracture at the time of follow-up. Another limitation could be the study size; however, we did find a substantial number of fractures in this population of patients. Last, the results might have been different if an orthopaedic surgeon had reviewed all of the images or if the neuroradiologist reviewed all of the images instead of only those with a discrepancy. Specifically, the utilization of an orthopaedic spine surgeon would have been most beneficial for detecting and classifying fractures in the present study.
In the present study, the nonreconstructed computerized tomographic scan outperformed standard radiographs when used to screen for thoracolumbar spine pathology. On the basis of our results, we propose that routine 5-mm collimation axial computerized tomographic images with supplemental routine thick-slab coronal (and/or sagittal) reformats are sufficient to diagnose all clinically important thoracic and lumbar spine injuries, with a high sensitivity for detecting most fractures. (The same results cannot be guaranteed when using a less-than-sixteen-slice scanner.) The nonreconstructed computerized tomographic scan did not miss any injury requiring operative or nonoperative therapy. We believe that the nonreconstructed computerized tomographic scan is a reliable "stand-alone" screening tool for thoracolumbar spine injuries, thus making standard radiographs of the spine unnecessary for the initial evaluation of trauma patients who undergo a computerized tomographic Pan-scan of the chest, abdomen, and pelvis. Moreover, reconstructions dedicated to the spine do not need to be ordered unless additional elucidation is required, whether on the basis of clinical suspicion, because of abnormal computerized tomographic scan findings, or for preoperative planning. Therefore, Advanced Trauma Life Support guidelines should be changed to fit the current imaging resources and literature. Accordingly, patients presenting with altered mental status do not need to undergo anteroposterior and lateral radiographs of the spine unless that body region (i.e., the chest) will not undergo a nonreconstructed computerized tomographic scan. Orthogonal planes to assess alignment are required and likely should include axial, coronal, and sagittal planes. Additional data are needed to appropriately weigh the costs (both in terms of monetary costs and in terms of radiation exposure) and benefits of completing a Pan-scan for all blunt trauma patients presenting with altered mental status. 