RADIOGRAPHIC SENSITIVITY AND NEGATIVE PREDICTIVE VALUE FOR

ACUTE CANINE SPINAL TRAUMA

JENNIFER KINNS, WILFRIED MAI, GABRIELA SEILER, ALLISON ZWINGENBERGER, VICTORIA JOHNSON, ANA CACERES,
ALEJANDRO VALDE S-MARTINEZ, TOBIAS SCHWARZ

The objectives of this study were to establish the sensitivity and negative predictive value of radiography for
acute spinal osseous lesions in the canine trauma patient, and to evaluate the interobserver variability in
radiographic assessment of the spine in traumatized dogs. This was a retrospective multiple observer blinded
study. The study population included 30 canine patients that presented following acute trauma, with clinical
signs attributable to the spinal column. Radiography and computed tomography (CT) were performed in all
cases. Radiographic interpretation was performed independently by four observers with different experience
levels who were blinded to clinical information (other than trauma) and the CT results. CT studies were
interpreted by a further three radiologists who formed a consensus opinion on the presence of specific osseous
lesions. Using the CT results as a gold standard, the sensitivities and negative predictive values of radiography
for specific osseous lesions were calculated. Interobserver agreement was also evaluated. Radiography was
found to have only a moderate sensitivity for fractures (72%) and subluxations (77.5%). Low negative predictive values were found for the presence of vertebral canal narrowing (58%) and fracture fragments within the
vertebral canal (51%). Interobserver agreement was only moderate to fair for most lesion types. In conclusion,
radiography cannot be used to reliably rule out potentially unstable acute vertebral lesions in the canine trauma
patient, and further imaging is therefore often indicated in the patient with a high risk of such injuries.
Veterinary Radiology & Ultrasound, Vol. 47, No. 6, 2006, pp 563570.

Key words: CT, dog, fracture, radiograph, radiology, spine, vertebra.

Introduction

of spinal cord trauma. A three-compartment model used to

classify human spinal trauma has been adapted for use in
small animals.1,2 Each vertebra is divided into a dorsal,
middle, and ventral compartment (Fig. 1). The dorsal
compartment is composed of the articular processes, laminae, pedicles, spinous processes, and supporting soft-tissue
structures. The middle compartment includes the dorsal
longitudinal ligament, the dorsal aspect of the annulus
brosus of the intervertebral disc and dorsal part of the
vertebral body. The ventral compartment contains the rest
of the vertebral body, the lateral and ventral annulus
brosus, the nucleus pulposus, and the ventral longitudinal
ligament. If two of the three compartments are damaged, a
fracture is considered unstable.
Diagnostic imaging plays a key role in the evaluation of
spinal trauma patients. Radiographs provide an inexpensive and rapid means for initial evaluation of the spine
following trauma, but they have limitations. Accurate positioning is essential in obtaining diagnostic quality spine
radiographs and sedation is usually required to achieve this
goal.36 In the trauma patient, however, sedation can release the muscular bracing that protects the spine from
subsequent damage. It is therefore difcult to obtain a diagnostic study without risking further injury. Consequently
many studies are obtained without sedation and are therefore limited by obliquity, and by muscular bracing that

CUTE SPINAL TRAUMA can cause unstable spinal injuries. Such injuries have the potential to lead to failure
of the vertebral column to protect the spinal cord and
nerve roots from severe insult. Resultant myelopathy and
radiculopathy can lead to temporary or permanent paralysis. Rapid and accurate evaluation of the vertebral column and the status of the spinal cord facilitate appropriate
treatment and a better informed prognosis.
Accurate evaluation of any spinal trauma for fracture
stability and vertebral dislocation is important for surgical
planning and prognosis. Unstable lesions have a high risk

From the Department of Clinical Studies, University of Pennsylvania

School of Veterinary Medicine, 3900 Delancey Street, Philadelphia, PA
19104-6010, USA (Kinns, Mai, Seiler, Caceres, Valdes-Mart nez), the Department of Surgical and Radiological Sciences, School of Veterinary
Medicine, University of California, Davis, CA 95616 (Zwingenberger), the
Radiology Department, Murdoch University Veterinary Hospital, Murdoch, 6150 Western Australia (Johnson), and the Department of Surgical
Sciences, School of Veterinary Medicine, University of Wisconsin-Madison, Madison, WI 53706 (Schwarz).
This paper was presented at the annual scientific meeting of the American College of Veterinary Radiology, Chicago, November 28December
3, 2005.
Address correspondence and reprint requests to Dr. Kinns, at the above
address. E-mail: jkinns@vet.upenn.edu
Received March 30, 2006; accepted for publication May 17, 2006.
doi: 10.1111/j.1740-8261.2006.00186.x

563

564

2006

KINNS ET AL.

interobserver variability in the interpretation of these

radiographs.

Materials and Methods

Medical records of the Matthew J. Ryan Veterinary
Hospital of the University of Pennsylvania were searched
to identify canine patients examined between 1999 and
2005, with a history of trauma and with clinical signs suggestive of injury to the spinal column. Only patients that
had a CT within 48 h of survey radiographs, and for which
both studies were available, were included in the study.

Radiographs

Fig. 1. Transverse computed tomography image of a normal seventh

cervical canine vertebra wherein the location of the three compartments used
in analysis is dened.

may mask clinically signicant subluxation. It is also not

possible to evaluate the spinal cord using survey radiography. While myelography can be used to evaluate cord
compression, this technique gives little information about
spinal cord hemorrhage or the potential for future insult
from unstable lesions. There is also an increased risk in
positioning a patient with an unstable spine for cervical or
lumbar injection of contrast medium.
Cross-sectional imaging techniques overcome many of
the limitations of survey radiography. Magnetic resonance
(MR) imaging and later generation computed tomography
(CT) systems allow multiplanar evaluation. CT is highly
sensitive to the presence of bone lesions and is the modality
of choice as a rst line approach to the human polytrauma
patient.711 CT has a sensitivity of up to 100% for acute
osseous lesions.11 It is also possible to perform a CT study
with the anesthetized or sedated patient strapped to a
spinal board, thus protecting against further damage. MR
imaging, where available, is the rst choice technique for
evaluation of the spinal cord and integrity of supporting
soft tissues but a signicant number of spinal fractures can
be missed.9,10
With the wider availability of cross-sectional imaging,
such as CT, the clinician and radiologist have to make a
decision whether radiographs are adequate as a screening
tool, or if advanced imaging should be recommended in
any patient with a high risk of vertebral lesions.
The objectives of this retrospective study were to evaluate the sensitivity and negative predictive value of radiography for acute osseous lesions in canine spine trauma
patients, using CT as a gold standard, and to measure

Most studies included lateral and ventrodorsal radiographs of the spinal region of interest, as localized by neurologic evaluation. Several studies were composed of
survey radiographs of the entire spine; these were typically patients in which multiple lesions were suspected clinically or where neurologic localization could not be
established. There were, however, a number of studies in
which the ventrodorsal view was not obtained due to concerns about spinal column instability. Lumbar myelography was performed in ve patients, and both survey and
myelographic views were included in the study.

CT
CT was performed using a helical single-slice CT unit
(GE ProSpeed). A typical study included a 3 mm slice
thickness and interval (3
3 mm) detail algorithm series of
the clinical region of interest followed by a 1
1 mm bone
algorithm series of any area of concern. All studies were
performed using nonhelical acquisition. The slice thickness
varied depending on the area of coverage and size of the
dog.

Data Collection
Radiographs from each patient were reviewed by four
radiologists who had different levels of experience. Two (A
and B) were board certied radiologists, one (C) was a
board eligible radiologist and one (D) was a radiology
resident. Each viewed the radiographs independently and
in random order during an unlimited time period. They
were aware that each patient had sustained acute trauma
with clinical signs attributable to a spinal lesion, but were
given no further clinical information.
A custom-designed form (Table 1) was used by the observers to record their ndings in each patient. This required that the observers answered yes or no to a
series of questions. For each yes or no answer they
were also required to select a degree of certainty, for which
General Electric Company, Milwaukee, WI.

Vol. 47, No. 6

565

SPINAL TRAUMA IN DOGS

Table 1. Summary of the Radiographic Evaluation Questionnaire

Question
List views available
Radiographs considered to be of diagnostic quality?
Presence of a vertebral fracture?
Which vertebrae are fractured?
Ventral compartment fractured?
Middle compartment fractured?
Dorsal compartment fractured?
Fracture fragments in the vertebral canal?
Luxation or subluxation present?
Which vertebrae are on either side of the lesion?
Vertebral canal narrowing?
Where are the sites of narrowing?

the categories were certain, fairly sure, or unsure.

With every fracture or luxation/subluxation identied they
were additionally required to list the location of the lesion,
and to answer any subsequent questions (such as compartment involvement) for each individual lesion. The
observers were given a description and diagram of the three
compartment model to which they were able to refer at
any time.

CT Evaluation
CT evaluation was carried out in a similar way to establish a gold standard for the study. The CT studies were
reviewed by two board certied radiologists and a radiology resident (all different from the radiograph reviewers)
who formed a consensus opinion as to the diagnosis in each
patient. A similar custom-designed form was used to record the CT ndings. Medical records were searched for
surgical and necropsy conrmation of diagnosis.

Statistical Analysis
Accuracy matrices were used to calculate the sensitivity
of radiographs to the presence of specific acute spinal osseous lesions, using the CT results as a gold standard.
Sensitivity was calculated for each observer and as a mean
value, for each lesion type. With regard to the recognition
of the presence of a fracture, sensitivity was calculated
under two conditions: under the rst condition (a) a true
positive included those patients in which any of multiple
fractures were recognized; under the second condition (b) a
true positive only constituted those patients in which all
fractures were identied. Analysis of vertebral compartment involvement, and recognition of fracture fragments
within the vertebral canal, only included those patients in
which a fracture had already been recognized by that observer. A negative predictive value was also calculated for
most radiographic parameters evaluated. In addition, a w2
test of independence was used to evaluate the correlation
between CT signs of cord compression and radiographic

Required Answer

Degree of Certainty

Radiographic views
Yes/No
Yes/No
List vertebrae
Yes/No
Yes/No
Yes/No
Yes/No
Yes/No
List vertebrae
Yes/No
List vertebrae

Not applicable
Certain/fairly sure/unsure
Certain/fairly sure/unsure
Not applicable
Certain/fairly sure/unsure
Certain/fairly sure/unsure
Certain/fairly sure/unsure
Certain/fairly sure/unsure
Certain/fairly sure/unsure
Not applicable
Certain/fairly sure/unsure
Not applicable

evidence of vertebral canal narrowing. The level of signicance was set at Po0.05.
k statistics were used to measure interobserver agreement. Each answer was assigned a value: nocertain
1, nofairly sure 2, nounsure 3, yesunsure
3, yesfairly sure 4, yescertain 5. With regard
to recognition of the presence of a fracture, interobserver
agreement could only be calculated under the rst condition
(a). k statistics were calculated across all observers, and
between experienced observers only (removing observer D)
for all lesion categories.
All statistical analysis was performed using a commercial
statistical software package.w

Results
Thirty dogs met the inclusion criteria. The mean age at
presentation was 4.8 years (median 4.5 years, range 6
months14 years). Breeds included mixed breed (7), Labrador Retriever (4), Beagle (3), Great Dane (2) and one
each of 14 other breeds. Seven intact and 12 neutered
males, and four intact and six neutered females were included. The majority of patients (21) had been hit by a car,
and three in which the origin of trauma was not conrmed
were also considered likely to have suffered a road trafc
accident. Five dogs fell from a height, and the remaining
dog ran into a tree.
CT demonstrated the presence of a variety of lesions, as
summarized in Table 2. Twenty-four dogs (80%) had acute
osseous lesions, of which 23 had vertebral fractures, nine
had vertebral subluxation with fracture, and one had subluxation only. There were no complete luxations. Patients
had vertebral fractures distributed fairly evenly between
the cervical (C), n 7, thoracic (T), n 9, and lumbar (L),
n 8, spine with one dog having a fracture in both the
cervical and thoracic spine. Fractures occurred with greater
frequency (64%) at the junction between mobile and immobile parts of the spine (C1C2, C7T1, T12L1, and
wIntercooled Stata 8.0 for Windows, College Station, TX.

566

2006

KINNS ET AL.

Table 2. Distribution of Lesions in 30 Dogs with Acute Spinal Trauma

Lesion Type

Number of Dogs

Fracture
Fracture and subluxation
Fracture and disc extrusion
Subluxation
Disc extrusion
Dural tear
Subdural hemorrhage
No lesion evident on CT

13
9
1
1
2
1
1
2

CT, computed tomography.

L7S1) than fractures occurring in a mid-segment location,

but the difference was not signicant (P 0.2). The CT
ndings were conrmed in eight patients that had surgery,
and in two further patients that had necropsy. The remaining patients did not undergo complete evaluation (either surgical or necropsy).
The radiographs were considered to be of diagnostic
quality by all observers in all but one dog. In this dog, one
observer considered that the views available were not adequate to complete the evaluation; this dog was therefore
excluded from the interobserver analysis. It was still included in the evaluation of radiographic sensitivities for
those observers that completed the examination for that
dog.
The sensitivities and negative predictive values of radiographs to each lesion type, including the value for each
observer and the mean value for all observers are summarized in Table 3. Specicities are not presented due to
the low number of true negative radiographs in the study.
Under condition a, sensitivity of radiography for the presence of a fracture had a mean of 82%. Under condition b,
72% mean sensitivity was found (Fig. 2), with a wide range
(54.582%) across observers. The negative predictive value
for recognition of all fractures was 49%. Among recognized fractures, differentiation into compartment
involvement resulted in a better sensitivity for ventral

Fig. 2. One-year-old Miniature Schnauzer with pelvic limb paralysis after

being hit by a car. (A) Lateral radiograph obtained at presentation. (B)
Transverse computed tomography image of the fth thoracic vertebra (detail
algorithm, wide window). There is a comminuted fracture involving all three
compartments. This fracture was not recognized radiographically by any
observer. An additional second cervical fracture was present that was recognized by two observers.

compartment fractures (92%) than middle or dorsal compartment fractures (83% and 84%, respectively). The presence of fracture fragments within the vertebral canal was
recognized with a mean sensitivity of 57.5% and negative
predictive value of 58%.
Seventeen (of 32) individual fractures were missed in
total by the observers, of which seven (41%) involved more
than one compartment, and were therefore unstable

Table 3. Summarized Results for the Sensitivity of Radiographs to Specific Aspects of Canine Spine Trauma
Sensitivity for Each Observer (%)
Type of Osseous Lesion
Presence of a fracture a
Presence of a fracture bw
Dorsal compartment fracturedz
Middle compartment fracturedz
Ventral compartment fracturedz
Presence of a subluxation
Fracture fragments within the vertebral canal
Vertebral canal narrowing

Mean (%)

95% CI (%)

Mean NPV (%)

95.5
82.0
95.0
95.0
86.0
90.0
69.0
65.0

95.5
72.5
82.5
78.0
100
90.0
61.5
65.0

95.5
77.5
82.0
89.5
89.0
80.0
67.0
60.0

72.5
54.5
75.0
69.0
92.5
50.0
33.0
40.0

90.0
72.0
84.0
83.0
92.0
77.5
57.5
58.0

71108
5291
7097
64101
82101
47107
3184
3976

76.5
49.0
46.5
61.0
81.0
88.0
58.0
51.0

True positives include those cases in which only one of multiple fractures was identied. wTrue positives only include cases in which all the fractures

present were identied. zDorsal, middle, and ventral compartment involvement was only evaluated in fractures that were recognized. CI, condence
interval; NPV, negative predictive value.

Vol. 47, No. 6

567

SPINAL TRAUMA IN DOGS

Fig. 3. Six-year-old mix breed dog with pelvic limb paresis that had been hit by a car. (A) Ventrodorsal radiograph of the thoracolumbar junction. (B)
Dorsal plane reconstruction image of a 1 mm thickness and increment computed tomography series of the 12th and 13th thoracic vertebrae, presented with a
wide window, showing compression secondary to vertebral canal narrowing with fracture fragments impinging on the compressed spinal cord. Neither vertebral
canal narrowing, nor fracture fragments within the vertebral canal, was recognized radiographically by any observer.

according to the three-compartment classication. One of

these fractures, involving the dorsal compartment only,
was missed by all observers. The experienced observers (A,
B, and C) missed nine fractures in total of which two (22%)
were unstable, four (44%) involved the dorsal compartment only and three (33%) involved the ventral compartment only.
Radiographic sensitivity to the presence of vertebral
canal narrowing (Fig. 3) had a mean value of 58%, and the
negative predictive value for vertebral canal narrowing was
51%. Vertebral canal narrowing was present on CT in 21
patients and resulted in spinal cord compression in 19
(90%) of them. Radiographic evidence of vertebral canal
narrowing (without myelogram) had a negative predictive
value of 35% for the presence of spinal cord compression.
Subluxations were recognized with a mean sensitivity of
77.5%. In one dog CT demonstrated the presence of a
subluxation that had not been recognized by any observer
radiographically (Fig. 4).
The results for interobserver agreement are summarized
in Table 4. Categories of agreement are based on an established classication.12 Moderate agreement was found
between observers in the recognition of the presence of a
fracture and in the identication of ventral compartment
involvement. Fair agreement was found for the presence of
subluxations, middle compartment involvement, and vertebral canal narrowing. Slight agreement was found across
observers for the identication of dorsal compartment involvement and for the recognition of fracture fragments
within the vertebral canal. k value was increased when
calculated across experienced observers only (A, B, and C)
for recognition of the presence of a fracture, identication

of middle compartment involvement, recognition of the

presence of a subluxation, recognition of fracture fragments within the vertebral canal and recognition of vertebral canal narrowing. Nonetheless, the category of
agreement (fair or moderate) was unaffected by removal
of the least experienced observer.

Discussion
The low negative predictive value (48%) and moderate
sensitivity (72%) for radiographic detection of all vertebral
fractures in an acute trauma patient suggests that radiography is inadequate to rule out acute osseous lesions.
Similar limitations have been found in humans, where sensitivities of 45%, 65%, and 93% are reported for the
radiographic detection of acute osseous lesions in cervical
trauma.11,13,14 The sensitivity in our study was better
(82%) if a true positive included dogs in which any of
several fractures was recognized, but this is not clinically
relevant as it could lead to mistaken treatment or prognosis. A prospective study of thoracolumbar trauma in
human patients suggested that, although radiography often
(12.5%) misclassied spinal lesions as chronic rather than
acute, unstable lesions were rarely missed.15 Nevertheless,
in our study 41% of fractures that were missed involved
more than one compartment and were therefore considered
unstable according to the three-compartment model.1
Several of the radiographic series in this study had been
obtained with suboptimal patient positioning. This study
was intended to replicate the situation in which a radiographic diagnosis is required in an emergency trauma patient; for this reason it was considered that these studies

568

2006

KINNS ET AL.

Fig. 4. Three-year-old American Pit Bull Terrier with pelvic limb paresis, having been hit by a car. (A) Lateral radiograph of the cranial aspect of the lumbar
spine. There is narrowing of the intervertebral disc space between the third and fourth lumbar vertebrae (L3L4). (B) Sagittal reconstruction of a 1 mm
thickness and increment axial CT series at L3L4, there is moderate subluxation at that site. Subluxation in this dog was not recognized radiographically by any
observer.

should be included. In humans, survey radiographs in cervical trauma patients were considered adequate (proper
positioning, correct exposure, and adequate visualization
of anatomy) by the radiologist in only 48% of patients,
also reecting the difculty in positioning acute trauma
patients.14
Fractures were correctly classied as involving the ventral compartment with greater sensitivity (92%) than involvement of the middle and dorsal compartments (83%
and 84% sensitivity respectively). However, of the fractures
that were missed, approximately one-third involved the
ventral compartment only, and it is therefore not possible
to conclude from these data that ventral compartment
fractures are more easily recognized.
Radiography in this study had a low negative predictive
value (58%) for the presence of fracture fragments within
the vertebral canal. Fragments that penetrate the spinal
cord are likely to cause severe local damage, while free
fragments present an ongoing risk for further spinal cord

or nerve root trauma and require surgical intervention.

This study suggests that radiography cannot be used to
rule out the presence of fracture fragments within the vertebral canal.
The sensitivity of radiography to the presence of subluxations was also found to be only moderate (75%).
Muscular bracing in the unsedated patient may have
masked the presence of subluxations that were then apparent under general anesthesia. It is unlikely that further
injury was caused during positioning for CT, as many of
the patients were imaged strapped to a spinal board.
The low negative predictive value (35%) for spinal cord
compression found here indicates that radiographic absence of vertebral canal narrowing cannot be used reliably
to indicate that there is no spinal cord compression. Although the neurologic status of the patient gives some indication as to whether a spinal cord injury is present, it is
important to be able to differentiate compressive (and
therefore surgical) lesions from those resulting in cord

Table 4. Interobserver Agreement in the Radiographic Evaluation of Specific Aspects of Spinal Trauma

Presence of a fracture
Dorsal compartment involvement
Middle compartment involvement
Ventral compartment involvement
Fracture fragments within the vertebral canal
Presence of a subluxation
Vertebral canal narrowing

k Value for Agreement

All Observers

k Value for Agreement

Experienced Observers

0.43
0.14
0.32
0.41
0.15
0.26
0.23

o0.001
0.01
o0.001
o0.001
0.002
o0.001
o0.001

0.59
0.14
0.36
0.41
0.24
0.27
0.31

o0.001
0.03
o0.001
o0.001
o0.001
o0.001
o0.001

Vol. 47, No. 6

SPINAL TRAUMA IN DOGS

swelling or hemorrhage. In people it has been shown that

compressive or unstable traumatic cervical lesions missed
radiographically can often result in permanent neurologic
deficits.16
There was a wide range of calculated sensitivities across
observers, and the differences mostly paralleled the degree
of experience, as one would expect. We decided to include
inexperienced observers in our analysis to reect the reality
of emergency practices where radiographic reports are
often nalized by an experienced radiologist only after a
clinical decision has been made.
Analysis of agreement across observers suggests that
radiographic diagnosis may be unreliable in dorsal and
middle compartment involvement, fracture fragments within the vertebral canal, subluxation, and vertebral canal
narrowing. Clinically this could lead to misinformed decisions based on the perceived stability and severity of a
lesion.
CT has been shown to be valuable in the investigation of
acute spinal trauma in dogs.6,17 While the sensitivity of CT
to the detection of acute osseous lesions has not been reported in dogs, the surgical and necropsy results that were
available in a minority of our dogs conrmed the CT
diagnoses. Data from human trauma patients also indicate
that the sensitivity of CT is very high. Values of between
97% and 100% have been reported in large prospective
clinical studies.10,11 CT is accepted as the gold standard in
imaging osseous spinal lesions in people. In a prospective
study of 688 trauma patients with cervical lesions where
MR and CT were compared for evaluation of the cervical
spine MR imaging allowed detection of only 55% of fractures while CT allowed detection of 97%.10
The types of lesions observed in this study were similar
to those seen in other reports of canine spine trauma.1823

569

In this study, lesions were distributed evenly between the

spinal segments, contrary to previous reports which indicated that the lumbar spine is the most common site of
injury.19,20,23 Our study is biased in selecting animals that
had CT evaluation; it may be that those with lumbar lesions were less likely to be referred or to have a CT. We did
not nd a signicant predilection for lesions to occur at
junctional locations, although this has been reported
previously.1820
It is recognized that the decision to perform a CT in
many of these patients was based on initial radiographic
ndings. This biased the study toward patients with radiographically apparent lesions, and as such has likely caused
us to overestimate the sensitivity of radiography to the
presence of fractures and luxations. It is also possible that
this population included patients with more severe neurologic disease, regardless of appearance on radiographs, and
this is also likely to bias the population toward patients
with unstable or compressive lesions. As this was a retrospective study, we also had no control over the specific CT
study that was performed in each dog. In larger dogs the
CT series chosen may have encompassed only a region of
concern identied radiographically, rather than a region of
clinical concern (such as T3L3). As such, additional acute
osseous may have been missed on CT, also leading to an
over estimation of radiographic sensitivity.
In conclusion, the results of our study suggest that radiography is inadequate to rule out vertebral fractures and
subluxations in the acute canine spinal trauma patient.
Radiography is also a poor diagnostic tool to assess the
stability of spinal fractures, and cannot reliably be used to
detect compressive lesions. CT, where available, is therefore recommended in patients with a high clinical suspicion
of such injury.

REFERENCES
1. Shores A, Braund KG, Brawner WR Jr. Management of acute
spinal cord trauma. Vet Med 1990;85:724732, 734739.
2. Shores A. Spinal trauma-pathophysiology and management of
traumatic spinal injuries. Vet Clin North Am Small Anim Pract 1992;
22:859888.
3. Lewis RE. Roentgen signs of the spine. Vet Clin North Am Small
Anim Pract 1974;4:647661.
4. Sande RD. Radiography, myelography, computed tomography, and
magnetic resonance imaging of the spine. Vet Clin North Am Small Anim
Pract 1990;22:811831.
5. Dennis R. Radiographic examination of the canine spine. Vet Rec
1987;121:3135.
6. Bagley RS. Spinal fracture or luxation. Vet Clin North Am Small
Anim Pract 2000;30:133153.
7. Blackmore CC, Emerson SS, Mann FA, Koepsell TD. Cervical spine
imaging in patients with trauma: determination of fracture risk to optimize
use. Radiology 1999;211:759765.
8. Mann FA, Cohen WA, Linnau KF, Hallam DK, Blackmore CC.
Evidence-based approach to using CT in spinal trauma. Eur J Radiol
2003;48:3948.

9. Tins BJ, Cassar-Pullicino VN. Imaging of acute cervical spine injuries: review and outlook. Clin Radiol 2004;59:865880.
10. Holmes JF, Mirvis SE, Panacek EA, Hoffman JR, Mower WR,
Velmahos GC. Variability in computed tomography and magnetic resonance imaging in patients with cervical spine injuries. J Trauma 2002;53:
524529.
11. Griffen MM, Frykberg ER, Kerwin AJ, et al. Radiographic clearance of blunt cervical spine injur: plain radiograph or computed tomography
scan? J Trauma 2003;55:222226.
12. Kundel HL, Polansky M. Measurement of observer agreement. [see
comment]. Radiology 2003;228:303308.
13. Nguyen GK, Clark R. Adequacy of plain radiography in the diagnosis of cervical spine injuries. Emerg Radiol 2005;11:158161.
14. McCulloch PT, France J, Jones DL, et al. Helical computed tomography alone compared with plain radiographs with adjunct computed
tomography to evaluate the cervical spine after high-energy trauma. J Bone
Jt Surg Am 2005;87:23882394.
15. Hauser CJ, Visvikis G, Hinrichs C, et al. Prospective validation of
computed tomographic screening of the thoracolumbar spine in trauma. J
Trauma 2003;55:228234.

570

KINNS ET AL.

16. Davis JW, Phreaner DL, Hoyt DB, Mackersie RC. The etiology of
missed cervical spine injuries. J Trauma 1993;34:342346.
17. Steffen F, Flueckiger M, Montavon PM. Traumatic atlanto-occipital
luxation in a dog: associated hypoglossal nerve deficits and use of 3-dimensional computed tomography. Vet Surg 2003;32:411415.
18. Patterson RH, Smith GK. Backsplinting for treatment of thoracic
and lumbar fracture/luxation in the dog: principles of application and case
series. Vet Comp Orthop Trauma 1992;4:179187.
19. McKee WM. Spinal trauma in dogs and cats: a review of 51 cases.
Vet Rec 1990;126:285289.

2006

20. Feeney DA, Oliver JE. Blunt spinal trauma in the dog and cat:
insight into radiographic lesions. J Am Anim Hosp Assoc 1980;16:885890.
21. Turner WD. Fractures and fracture-luxations of the lumbar spine: a
retrospective study in the dog. J Am Anim Hosp Assoc 1987;23:459464.
22. Hawthorne JC, Blevins WE, Wallace LJ, Glickman N, Waters DJ.
Cervical vertebral fractures in 56 dogs: a retrospective study. J Am Anim
Hosp Assoc 1999;35:135146.
23. Selcer RR, Bubb WJ, Walker TL. Management of vertebral column
fractures in dogs and cats: 211 cases (19771985). J Am Vet Med Assoc
1991;198:19651968.