| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Department of Radiology, University of Sydney, PO Box 11, Woden ACT 2606, Australia
Editor:
The recently published article by Katzberg and colleagues in the October 1999 issue of Radiology (1), while a worthy contribution to our knowledge about magnetic resonance (MR) imaging of cervical spine trauma, fails to tackle the fundamental questions that must be answered for this modality to supplant conventional radiography in acute trauma imaging.
On the basis of the five-level hierarchy of imaging modality assessment (2), their work merely addresses the first level, imaging effectiveness, where the imaging modality in question, MR imaging, is compared with a previously available technique, conventional radiography. The value of MR imaging, in comparison with radiography, has been known for some time (3). Despite Katzberg and colleagues’ article having too few patients for comparison with computed tomography (CT), the role of CT imaging of acute cervical trauma has also been established (4).
The second level, diagnostic effectiveness, which involves assessment of a modality against clinical findings (such as the ability of MR imaging to aid in the confirmation of the disease accompanying a neurologic deficit), was not assessed in their study, in contrast to work by Flanders et al (4) and by me and my colleagues (5).
The third level, therapeutic effectiveness, which involves an assessment of the influence of a technique on patient treatment, has also been assessed previously (5,6) but was not addressed by Katzberg and colleagues.
The fourth level, the effect on patient outcome, has been only partially addressed in several studies, including our own (5) and that of Flanders et al (6), but was also not assessed by Katzberg and colleagues.
Finally, the fifth level, cost-effectiveness, which perhaps has the most relevance when determining whether an expensive technique (MR imaging) can supplant a relatively inexpensive one (conventional radiography), to my knowledge has yet to be assessed by anyone.
Katzberg and colleagues’ study certainly adds to the body of knowledge in this area, but considerably more work is required before MR imaging can truly supplant radiography in the imaging of acute cervical trauma.
REFERENCES
,
Christiana M. Drake, PhD,
,
Marija Ivanovic, PhD,*,
Richard A. Levine, PhD,,
Carol S. Beatty, MD,*,
William R. Nemzek, MD,*,
Russell A. McFall, MD,*,
Francesca K. Ontell, MD*,
Virginia C. Poirier, MD,§ and
Brian W. Chong, MD||
Departments of Radiology* and Biostatistics, University of California-Davis Medical Center, 4701 X Street, Sacramento, CA 95817; Southern Oregon Imaging, Medford Radiological, Medford, Oregon, Aurora, OH;§ Department of Radiology, University of Utah School of Medicine, Salt Lake City, UT||
We thank Dr Silberstein for his comments regarding our article "Acute Cervical Spine Injuries: Prospective MR Imaging Assessment at a Level I Trauma Center" published in Radiology (1). We are pleased to help clarify the issues he raised surrounding clinical imaging research such as ours.
As clearly stated in the referenced article, our focus was solely on detection of injury at MR imaging and at radiography. We believe there is still a void in this step of Dr Silberstein’s referenced five-level hierarchy of imaging assessment (2). We disagree that the value of MR imaging in comparison with radiography and CT has been known for some time. Our reasons for this are as follows.
All prior studies, to our knowledge, have been retrospective and highly selective toward patients with injury. In previous retrospective studies, such as those by Kulkarmi et al (3), 19 (70%) of 27 patients had cord abnormalities, and 21 (78%) of 27 patients had skeletal or ligamentous injuries. In the study by Mirvis et al (4), all 21 patients had acute neurologic deficiencies following cervical spine trauma. In the study by Flanders et al (5), all 78 patients had a demonstrable cervical spine fracture or subluxation or a neurologic deficit. In the study by Orrison et al (6), in which they compared MR imaging and CT and/or radiography, 77 (68%) of 113 patients had abnormalities.
Dr Silberstein uses his own reference (7) to support the assertion that "the value of MR, in comparison with radiography, has been known for some time." It is notable that his study was also retrospective, and all 34 patients had injuries. This considerable problem of selection bias was highlighted in an editorial in Radiology by Hackney (8) in discussion of Flanders et al stating that "the study by Flanders et al was of a highly selected subset of patients with spinal injury" and "probably does not represent an unbiased sample of patients with spinal injuries."
A further weakness of all studies has been the absence of a rigorous standard of truth. Even in the important investigation by Flanders et al, truth was used as the reference standard in the comparison of CT with MR imaging, the modality reputedly offering the best depiction of a particular abnormality. Thus, CT was considered the reference standard in all observations that related to disruption of the bone axis, and MR imaging was the reference standard for all observations that related to the spinal cord and the paraspinal soft tissues. The study by Dr Silberstein and colleagues also has this weakness, since they stated, "the imaging modality which reputedly offered optimal identification of a particular abnormality was used as a standard of reference."
We attempted to minimize selection bias, gather patients prospectively, and establish a more comprehensive standard of truth. Thus, we established our standard of truth by having two experienced radiologists assess all radiographs, all MR images at the time of presentation or thereafter, all CT scans, emergency department clinical records, prehospital clinical records, inpatient records with the discharge summary, and surgical records, if surgery was performed. The final determination of the standard of truth was completed only after all the images and records were collated and assessed.
Our imaging assessments were accomplished prior to spinal injury reduction and depicted injury within 24 hours of presentation in 189 patients. Our readers were blinded to findings of other imaging studies, worked independently, and assessed each vertebral level from C1 through T1, which resulted in 70,048 distinct observations.
Our statistical assessment of significance used a detailed analysis of vertebral levels, and significance was accepted only when there was no overlap in the CIs between modalities. In circumstances in which CIs could not be calculated owing to the complete lack of the detection of a positive finding by means of one modality (such as the inability of conventional radiography to depict traumatic disk herniation, cord edema, or cord compression) it was assumed that MR imaging was significantly better if there were at least 10 positive observations with that modality. Thus, in the case of paravertebral hemorrhage or edema, conventional radiography depicted none of the occurrences in 32 patients at 141 levels, whereas MR imaging had a 60% (CI: 42%, 77%) weighted mean sensitivity for this observation. In this particular circumstance, we concluded that MR imaging was significantly better than conventional radiography. We used the 
2 test, with the patient as the unit of observation, to assess associations between injury types (such as cord edema) and risk factors (such as cord stenosis), with significance accepted when the P value was less than .05. Reader agreement for conventional radiography and MR imaging was measured by means of 
statistics (9).
We believe it is premature of us to accomplish levels two to five of Dr Silberstein’s referenced five-level hierarchy of imaging assessment. Although our initial database is enormous for its singular purpose, it is not adequately large to fully incorporate data about incidence of injury, which is necessary for the additional levels of study. Indeed, we believe that further work is still required to complete even the first step of the hierarchy of imaging assessment, since CT has still not been adequately incorporated into anyone’s study of injury detection.
We hope that other prospective studies such as ours can incorporate a more comprehensive analysis of the relative merits of CT, conventional radiography, and MR imaging in the initial presentation of patients with acute cervical injuries. We believe we have properly concluded in the abstract of our study (1) that "MR imaging is more accurate than radiography in the detection of a wide spectrum of neck injuries, and further study is warranted of its potential effect on medical decision making, clinical outcome, and cost-effectiveness." The relative imaging effectiveness of conventional radiography, MR imaging, and CT, or level one of the five-level hierarchy of imaging assessment, is still unknown.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
